# Quantum machine learning speedups compared to classical AI

The intersection of quantum mechanics and machine learning represents one of the most heavily theorized and rigorously debated frontiers in contemporary computer science and physics [cite: 1, 2]. By mapping classical or quantum-native data into the exponentially large vector spaces inherent to quantum systems and leveraging the physical phenomena of superposition, entanglement, and quantum interference, researchers hypothesize that Quantum Machine Learning (QML) could surpass classical machine learning in both computational time complexity and feature representation [cite: 3, 4, 5]. 

The central inquiry dominating the discipline is whether quantum computers offer genuine, mathematically provable speedups over classical artificial intelligence, or if the projected capabilities are primarily the result of idealized theoretical models that fail to account for severe hardware realities and highly optimized classical algorithms. An exhaustive analysis of academic literature, algorithm development, and hardware roadmaps between 2018 and 2026 reveals a highly nuanced landscape. While early claims of exponential algorithmic speedups for processing classical data have been largely systematically dismantled through a process known as "dequantization," new frameworks for processing quantum-native data and streaming classical data have successfully established mathematically rigorous domains of true quantum advantage [cite: 6, 7, 8, 9].

This report examines the underlying mathematical mechanisms of QML, the severe limitations imposed by classical data loading, the theoretical boundaries of algorithmic speedups, the trajectory of fault-tolerant hardware, and the geopolitical investment landscape funding these technological developments.

## Mathematical Foundations of Quantum Information Processing

At its core, classical machine learning operates by optimizing parameters across high-dimensional feature spaces to identify patterns, classify data, or generate predictive text [cite: 2, 4]. These models process information sequentially or in parallel using bits and floating-point arithmetic. QML diverges fundamentally from this paradigm by encoding information into quantum states, frequently represented by density matrices, with transformations executed via unitary operations (quantum gates) [cite: 1, 2]. 

A single quantum system containing $n$ qubits exists in a $2^n$-dimensional complex vector space, known as a Hilbert space. This mathematical property allows an $n$-qubit system to inherently represent an astronomically large feature space [cite: 5]. The theoretical appeal of QML relies heavily on the premise that this exponential dimensionality can be utilized to evaluate global properties of a dataset simultaneously through quantum parallelism, rather than evaluating data points sequentially [cite: 3, 10, 11].

QML models generally fall into distinct combinatorial categories based on the nature of the data being processed and the hardware executing the algorithm:

1.  **Classical-Classical (CC):** Standard machine learning using classical data on classical hardware.
2.  **Quantum-Classical (QC):** Quantum algorithms designed to process classical data. This is often termed "quantum-enhanced machine learning" and represents the vast majority of commercial enterprise applications [cite: 1, 12].
3.  **Classical-Quantum (CQ):** Classical machine learning applied to understand, optimize, or control physical quantum systems, such as optimizing quantum error correction codes or learning Hamiltonians [cite: 1, 13].
4.  **Quantum-Quantum (QQ):** Quantum algorithms processing quantum states directly generated from physical quantum systems, completely bypassing classical data translation [cite: 1, 13].

The vast majority of commercial interest, venture capital, and industry hype surrounds the QC domain—using quantum processors to accelerate classical tasks such as image recognition, natural language processing, and financial portfolio optimization [cite: 1, 12, 13]. However, this specific domain is precisely where the theoretical advantages of QML face the most severe algorithmic and physical bottlenecks.

## Algorithmic Speedups and the Dequantization Paradigm

For much of the 2010s, researchers proposed various QML algorithms that promised exponential speedups over their best-known classical counterparts [cite: 14]. A foundational example was the 2016 algorithm by Kerenidis and Prakash for recommendation systems, which mathematically demonstrated an exponential reduction in processing time compared to classical recommendation engines [cite: 6, 14]. This algorithm was widely considered one of the strongest candidates for demonstrating provable exponential speedups in machine learning [cite: 6].

This assumption of intrinsic quantum superiority was fundamentally challenged by the work of researcher Ewin Tang, initiated in 2018 and expanded upon through 2024. Tang successfully "dequantized" the Kerenidis-Prakash recommendation algorithm, proving that a classical randomized algorithm could achieve similar performance metrics with only a minor polynomial overhead, rather than an exponential slowdown [cite: 6, 14, 15]. 

### The Mechanism of Dequantization

Dequantization exposes a critical, often minimized assumption hidden within many early QML proposals: the reliance on Quantum Random Access Memory (QRAM) [cite: 9, 16]. Many QML algorithms achieve their exponential speedup by assuming that massive amounts of classical data can be mapped into a quantum superposition in logarithmic time, $O(\log N)$. To do this, theorists relied on an idealized QRAM structure that allows quantum computers to query classical databases entirely in superposition [cite: 16, 17].

Tang demonstrated that if a quantum algorithm is granted the powerful assumption of QRAM, a classical algorithm must be granted a mathematically equivalent data structure to ensure a fair complexity comparison. Specifically, Tang provided the classical algorithm with $l^2$-norm sampling access to the input data distributions [cite: 6, 15]. With this specific access, classical algorithms can utilize randomized sketching and Monte Carlo techniques to accurately mimic the quantum state's probability distribution [cite: 18, 19]. 

Following Tang's breakthrough, a generalized framework emerged—formalized by researchers including Chia, Gilyén, Li, Lin, Tang, and Wang—that successfully dequantized a wide array of QML algorithms operating on low-rank matrices [cite: 15, 18, 19]. The algorithms proven to lack exponential quantum advantage under these specific data-access assumptions include:

*   Quantum Principal Component Analysis (QPCA)
*   Quantum Support Vector Machines (QSVM)
*   Supervised clustering algorithms
*   Low-rank stochastic regression
*   Sublinear classical algorithms for solving low-rank linear systems

By utilizing stochastic techniques and exploiting special properties within the input data, these classical counterparts replicate the quantum output with minor polynomial overhead [cite: 15, 18, 20]. Consequently, the scientific consensus shifted drastically: for standard classical data represented as dense, low-rank matrices, QML does not offer an exponential algorithmic speedup if the classical counterpart is granted equivalent memory access [cite: 6, 15, 21]. 

| Algorithm | Original Quantum Promise | Dequantization Status | Classical Counterpart Performance |
| :--- | :--- | :--- | :--- |
| **Recommendation Systems** | Exponential speedup | Dequantized (Tang, 2019) | Polynomial slowdown using $l^2$-norm sampling [cite: 6, 14] |
| **Principal Component Analysis** | Exponential speedup via eigenvalue estimation | Dequantized | Randomized classical sketching matches performance [cite: 15, 18] |
| **Support Vector Machines** | Exponential speedup via quantum kernel estimation | Dequantized (for low-rank data) | Comparable complexity under QRAM-equivalent assumptions [cite: 15, 18] |
| **Supervised Clustering** | Exponential speedup | Dequantized | Polynomial equivalent established [cite: 15, 18, 19] |

## Classical Data Loading and State Preparation Bottlenecks

Even if a QML algorithm resists mathematical dequantization, it faces a severe physical barrier regarding the initialization of classical data into a quantum state. This is formally known in the literature as the state preparation or data loading bottleneck [cite: 4, 8, 16]. 

Before a quantum processor can manipulate classical data—such as a high-resolution image, a complex molecular structure, or a financial time series—the classical bits must be translated into the amplitudes, basis states, or phases of qubits. Given the diverse applications of QML, multiple encoding strategies have been developed, each with distinct scaling challenges:

| Encoding Method | Mechanism | Primary Limitation |
| :--- | :--- | :--- |
| **Basis Encoding** | Maps binary strings directly to computational basis states (e.g., classical '1001' becomes $|1001\rangle$). | Highly inefficient for continuous variables; requires $N$ qubits for $N$ bits [cite: 5]. |
| **Amplitude Encoding** | Normalizes continuous classical vectors and maps them to the continuous amplitudes of a quantum state. | Allows exponential compression (encoding $2^n$ values in $n$ qubits), but preparing the state natively requires an intractable $O(2^n)$ quantum gates [cite: 5, 22, 23]. |
| **Angle / Sinusoidal Encoding** | Encodes classical features as rotation angles of quantum gates. | Only loads one feature per qubit; fails to exploit the exponential Hilbert space for compression [cite: 11, 23]. |

For dense, high-dimensional data, basic encoding strategies like amplitude encoding require deep, complex quantum circuits that introduce overwhelming noise on current hardware [cite: 22, 23]. 

### The Impact of the Holevo Bound

Furthermore, the extraction of data from a quantum system is strictly restricted by the Holevo bound. The Holevo bound mathematically stipulates that while $n$ qubits can represent an exponentially large data state in internal superposition, the maximum amount of classical information that can be extracted via measurement is exactly $n$ bits [cite: 4, 22, 24]. 

This physical asymmetry creates a severe operational constraint for end-to-end QML applications:
1.  Loading $N$ classical data points into a quantum state takes linear time $O(N)$ or requires theoretical, currently unbuilt QRAM architectures [cite: 23, 25].
2.  Once successfully loaded, the quantum computer can process the high-dimensional space exponentially faster.
3.  Due to the Holevo bound, reading the exact results out into a classical format collapses the superposition, destroying the probability wave and limiting retrieval [cite: 4, 17, 22].

Because of these I/O bottlenecks, for many realistic scenarios in the Noisy Intermediate-Scale Quantum (NISQ) era, the physical time required to encode the data into the machine and extract the answer completely nullifies any computational speedup achieved during the quantum processing phase [cite: 16, 17, 23, 26].

[image delta #1, 0 bytes]





## Near-Term Heuristics and Noisy Hardware Constraints

The current generation of quantum hardware exists within the Noisy Intermediate-Scale Quantum (NISQ) era—a term coined by physicist John Preskill in 2018 to describe systems that possess between 50 and 1,000 physical qubits but lack the capability for large-scale quantum error correction [cite: 27, 28]. In this regime, qubits easily lose their delicate quantum properties (a process known as decoherence) when exposed to minimal environmental disturbances like heat, vibrations, or electromagnetic interference [cite: 29, 30].

Because deep algorithms cannot execute reliably on error-prone hardware, the prevailing approach for QML in the near-term relies on Variational Quantum Algorithms (VQAs) [cite: 31]. VQAs are hybrid quantum-classical algorithms. A Parameterized Quantum Circuit (PQC) prepares a quantum state using a shallow circuit depth; the state is then measured to evaluate a cost function, and a classical optimizer (such as gradient descent) updates the circuit's parameters iteratively [cite: 11, 27, 31]. Variational Quantum Classifiers (VQCs) and Quantum Neural Networks (QNNs) operate entirely on this principle, bypassing the need for deep circuits by offloading the heavy optimization tasks to classical CPUs [cite: 11, 31].

### The Barren Plateau Phenomenon

While theoretically promising for circumventing hardware noise, VQAs face a severe mathematical limitation known as "Barren Plateaus." In foundational research published by McClean et al. (2018), it was discovered that in high-dimensional parameterized quantum circuits, the gradient of the cost function vanishes exponentially as the number of qubits increases [cite: 31, 32, 33]. 

This "flatness" in the optimization landscape means that classical optimizers cannot find the minimum to update the weights, rendering large-scale VQAs effectively untrainable without highly specific, structured initializations [cite: 4, 31]. Due to these optimization failures, empirical benchmarks of NISQ-era algorithms often show underwhelming results compared to classical baselines. A comprehensive 2025 benchmark study comparing a range of variational quantum algorithms against classical machine learning models for time series forecasting across 27 prediction tasks found that quantum models consistently struggled to match the accuracy of simple classical counterparts of comparable complexity [cite: 34]. 

## Fault-Tolerant Algorithmic Architectures

To achieve the exponential speedups theorized by quantum mechanics, QML must transition from heuristic VQAs to deep, mathematically rigorous algorithms that require Fault-Tolerant Quantum Computing (FTQC) [cite: 4, 28, 35]. 

### The Harrow-Hassidim-Lloyd (HHL) Algorithm

Introduced in 2009, the Harrow-Hassidim-Lloyd (HHL) algorithm is the foundational bedrock of FTQC algorithmic theory for machine learning. It is designed to solve systems of linear equations of the form $Ax = b$ [cite: 21, 36]. 

Classically, finding the vector $x$ takes time polynomial in $N$ (the number of variables), typically scaling as $O(N\kappa)$ where $\kappa$ is the condition number of the matrix $A$. The HHL algorithm reduces this complexity to $O(\log(N)\kappa^2)$, providing a true exponential speedup in the dimension $N$ [cite: 21, 37, 38]. Because matrix inversion and linear algebra form the mathematical engine of machine learning optimization, HHL theoretically enables quantum support vector machines, quantum neural networks, and least-squares learners to operate on feature spaces too vast for classical supercomputers [cite: 36, 39].

However, the HHL algorithm possesses strict mathematical prerequisites that limit its immediate applicability:
1.  **Matrix Sparsity:** The input matrix $A$ must be sparse, containing relatively few non-zero entries per row [cite: 21, 37, 38].
2.  **Conditioning:** The matrix must be well-conditioned, exhibiting a low condition number $\kappa$ [cite: 21, 37].
3.  **State Preparation:** The vector $b$ must be efficiently loaded into the quantum state $|b\rangle$ without requiring $O(N)$ operations [cite: 17, 21, 37].
4.  **Output Extraction Limitations:** The algorithm outputs the solution as a quantum state $|x\rangle$. Measuring the state to determine every specific entry of $x$ would require repeating the algorithm $O(N)$ times, completely destroying the exponential speedup. Therefore, the user must only require aggregate scalar properties of $x$, such as an expectation value $\langle x|M|x\rangle$ for some operator $M$ [cite: 17, 21, 36].

While Tang's work successfully dequantized dense, low-rank systems, the exponential speedup of HHL remains mathematically robust for sparse, high-rank matrices where classical approximation algorithms fail [cite: 37, 38].

## Domains of Mathematically Proven Quantum Advantage

To separate industry hype from reality, it is necessary to identify where QML avoids the pitfalls of dequantization, barren plateaus, and classical data loading overheads. Recent academic and empirical breakthroughs highlight two specific domains of genuine quantum advantage.

### Learning from Quantum-Native Data

The most uncontested domain for QML is the analysis of quantum-native data—data generated directly from a quantum physical system, such as a molecule, a superconducting qubit, or a quantum material sensor. In these Classical-Quantum (CQ) or Quantum-Quantum (QQ) paradigms, the classical data loading bottleneck simply does not exist because the data is inherently formatted as a quantum state [cite: 7, 8].

A classical machine learning algorithm operating on quantum systems must first measure the system. This act of measurement destroys the quantum superposition, collapsing it into a classical "snapshot." To fully reconstruct a quantum state composed of $n$ qubits, classical algorithms require approximately $O(2^n)$ discrete experiments [cite: 7]. 

Conversely, a quantum machine learning algorithm can directly interact with the unmeasured quantum state. By utilizing techniques like Bell measurements—where two saved copies of a quantum state are entangled and measured simultaneously—the QML algorithm evaluates correlations directly without collapsing the state prematurely [cite: 1, 7]. In an experiment utilizing Google's Sycamore processor, researchers demonstrated that QML required only a linear $O(n)$ number of experiments. For a 20-qubit system, this mathematical scaling resulted in the QML algorithm requiring 10,000 times fewer measurements than classical models to reach a 70% prediction accuracy [cite: 7]. Crucially, this advantage is information-theoretic; it is permanent and cannot be matched by any future increase in classical computing power [cite: 7].

### Quantum Oracle Sketching (2026 Breakthrough)

While QML on classical data has long been hindered by QRAM assumptions, an April 2026 publication by a joint Caltech, Google Quantum AI, and MIT team introduced a novel framework titled "Quantum Oracle Sketching" that established a provable exponential space advantage for processing massive classical data [cite: 8, 9, 24].

The algorithm circumvents the theoretical QRAM requirement by processing classical data as a continuous stream. For each classical data sample observed, the algorithm applies a small, incremental quantum rotation and immediately discards the sample without storing it in memory [cite: 8, 9, 24]. These sequential rotations accumulate to incrementally approximate a target quantum oracle "on the fly." 

Coupled with a readout technique known as interferometric classical shadows, Quantum Oracle Sketching completely bypasses the data loading and Holevo measurement bottlenecks [cite: 24, 40, 41]. The research proved mathematically that a quantum computer of polylogarithmic size (requiring fewer than 60 logical qubits) could perform complex classification and dimension reduction on massive datasets, consuming four to six orders of magnitude less memory than the most efficient classical sparse-matrix algorithms [cite: 9, 41]. This advantage persists irrespective of computational complexity conjectures (i.e., even if BPP = BQP) and relies strictly on the mathematical constraints of the Born rule [cite: 9, 41, 42]. The team validated these advantages using real-world datasets, including single-cell RNA sequencing and IMDb movie review sentiment analysis [cite: 9, 41].

## Error Mitigation and Algorithmic Resilience

Given that fault-tolerant systems requiring millions of physical qubits are not yet available, researchers are actively developing Quantum Error Mitigation (QEM) techniques to extend the utility of current NISQ hardware. Unlike Quantum Error Correction (QEC), which uses redundant qubits to actively fix errors during computation, QEM involves the classical post-processing of quantum measurement outcomes to mathematically rectify noise-induced effects [cite: 29, 43].

Common data-driven QEM techniques include:
*   **Zero-Noise Extrapolation (ZNE):** Running quantum circuits at artificially elevated noise levels and mathematically extrapolating the results backward to predict a zero-noise outcome [cite: 29, 43].
*   **Probabilistic Error Cancellation (PEC):** Actively reversing specific profiles of quantum noise using mathematical models of the hardware's error channels [cite: 29].
*   **Virtual Distillation (VD):** Preparing multiple copies of a quantum state to identify and discard erroneous outcomes [cite: 43].

In 2023, researchers proposed unifying these disparate techniques under a single data-driven framework called UNITED (UNIfied Technique for Error mitigation with Data) [cite: 43]. Furthermore, a significant 2024 presentation by IBM researchers at the APS March Meeting detailed the integration of classical machine learning directly into the error mitigation pipeline (ML-QEM). By training classical models—including linear regression, random forests, multi-layer perceptrons, and graph neural networks—on device-noise profiles, researchers successfully mitigated errors on diverse quantum circuits containing up to 100 qubits [cite: 44, 45]. The ML-QEM approach drastically reduced the exponential computational overhead typically associated with standard mitigation techniques, yielding near noise-free results for quantum algorithms running on imperfect hardware [cite: 44, 45].

## Structural Constraints in Large Language Models

The unprecedented commercial and technical success of Large Language Models (LLMs) like OpenAI's GPT-4 and Meta's LLaMA has triggered widespread speculation regarding the quantum acceleration of generative artificial intelligence. However, technical consensus across the quantum industry firmly debunks the notion that quantum computers will replace classical GPUs for end-to-end LLM training in the foreseeable future [cite: 46, 47, 48, 49].

The limitations preventing quantum LLM training are severe and structural. Modern LLMs contain hundreds of billions, or even trillions, of trainable parameters. Because every quantum parameter requires a distinct representation within a delicate qubit state, running a production-scale LLM natively would demand billions of high-fidelity, coherent qubits [cite: 46, 48, 50]. Current quantum processors support only a few hundred to a thousand noisy physical qubits, which degrade due to environmental decoherence within fractions of a second [cite: 30, 46]. Furthermore, LLM architectures rely heavily on approximate vector additions and self-attention mechanisms that are already hyper-optimized for parallel execution on classical hardware, leaving little room for quantum parallelism to provide an edge [cite: 48].

### Hybrid Quantum-Classical Adaptations

Despite the impossibility of full native quantum LLM training, hybrid approaches show early, specialized viability for fine-tuning. In 2026, researchers at Multiverse Computing successfully demonstrated the integration of quantum capabilities into a production-scale classical LLM using block-diagonal unitary (BDU) adapters. 

Instead of attempting to move the entire massive network to a quantum processor, the research team inserted Cayley-parameterized unitary adapters specifically into a pre-trained classical 8-billion-parameter Llama 3.1 model. Executing this fine-tuning step on a 156-qubit IBM processor required only 6,000 additional quantum parameters to achieve a 1.4% improvement in perplexity [cite: 46, 50]. The experiment demonstrated that quantum layers could aid in capturing highly complex, non-local language correlations that classical architectures struggle to identify [cite: 46, 50]. Thus, while quantum computers will not train foundation models natively, quantum-enhanced architectures may be utilized as specialized classification heads or parameter-efficient fine-tuning modules to reduce overall energy consumption [cite: 46, 49, 51].

## Hardware Scaling and the Transition to Logical Qubits

The long-term viability of exponential QML speedups is inextricably linked to the progression of underlying quantum hardware. To execute deep, mathematically proven algorithms like HHL, hardware must transition from fragile physical qubits to stable logical qubits [cite: 30, 52]. Logical qubits utilize Quantum Error Correction (QEC) protocols to spread a single unit of quantum information across dozens or hundreds of physical qubits, detecting and correcting errors actively without measuring and destroying the underlying quantum state [cite: 28, 29, 30].

The hardware industry is currently undergoing the official transition from the NISQ era toward the FTQC era [cite: 4, 28, 35]. Milestones achieved across 2024 and 2025 validate this trajectory. Google Quantum AI’s "Willow" chip (105 qubits) successfully demonstrated that scaling physical qubits can reduce overall logical error rates, crossing the critical threshold necessary for eventual fault tolerance [cite: 28, 46].

IBM’s updated 2024 technology roadmap explicitly codifies the timeline for logical qubit scaling. Recognizing that raw physical qubit counts are meaningless without corresponding improvements in gate fidelity, IBM's roadmap projects the realization of quantum utility through modular architectures and advanced error correction [cite: 53, 54, 55].

[image delta #2, 0 bytes]

 

Key milestones in the transition to fault tolerance include:
*   **2024 (IBM Heron):** Execution of circuits containing 5,000 gates with the lowest error rates achieved to date [cite: 53, 55].
*   **2028 (IBM Flamingo):** Execution of circuits containing 15,000 gates using advanced error mitigation techniques [cite: 53, 54].
*   **2029 (IBM Starling):** Delivery of the first commercial fault-tolerant quantum computer, utilizing approximately 200 logical qubits capable of running 100 million gate operations [cite: 53, 54, 56].
*   **2033 (IBM Blue Jay):** Deployment of a large-scale FTQC system utilizing 2,000 logical qubits capable of executing 1 billion gates [cite: 53, 54, 57].



## Geopolitical Funding Mechanisms and Legislative Strategies

The realization of fault-tolerant QML architectures requires enormous, sustained capital expenditure. In recognition of the profound economic and national security implications—ranging from the decryption of cryptographic standards to the acceleration of sovereign artificial intelligence—governments and private entities have committed over $65.9 billion globally to quantum initiatives through 2026 [cite: 58]. 

The global quantum ecosystem is distinctly tri-polar, divided fundamentally by funding mechanisms and strategic intent:

### The United States
The US ecosystem is predominantly driven by aggressive private venture capital and large technology conglomerates (including Google, IBM, and Microsoft). In 2024, private global investment in quantum startups surged 50% year-over-year to nearly $2.0 billion, with the US capturing roughly 50% of this global capital flow [cite: 59, 60, 61]. Additionally, the first three quarters of 2025 alone saw $1.25 billion injected into quantum startups globally [cite: 62, 63]. Federal initiatives, such as the DARPA Underexplored Systems for Utility-Scale Quantum Computing (US2QC) program, seek to evaluate whether computational value can exceed hardware costs by 2033 [cite: 62]. Consequently, US companies dominate hardware commercialization and maintain a substantial lead in overall quantum computing patents granted globally [cite: 61, 64].

### China
China approaches quantum supremacy as a critical, state-directed imperative, executing the largest centralized investment strategy globally. The Chinese government has committed an estimated $15 billion toward quantum technologies, significantly outpacing the estimated $6 billion in US federal funding [cite: 58, 63]. In March 2025, China announced the establishment of a 1 trillion RMB ($138 billion) venture capital guidance fund targeting deep tech, including quantum technologies and artificial intelligence [cite: 58, 62, 63]. Unlike the US, China's research focus heavily skews toward quantum communications and security networks—representing 39% of global publications in the field—to establish physically unhackable state infrastructure and sovereign supply chains [cite: 58, 61, 64].

### Europe and the 2026 Quantum Act
Europe generates the highest concentration of quantum research publications and talent globally, driven by academic institutions and steady public funding. However, the continent has historically failed to translate this academic excellence into commercial dominance [cite: 60, 65]. In 2024, despite housing roughly 32% of global quantum companies, Europe attracted only 5% of global private quantum capital [cite: 59, 60, 66].

To address deep market fragmentation across its member states and combat severe supply chain vulnerabilities, the European Commission developed the Quantum Europe Strategy in 2025, paving the way for the European Quantum Act [cite: 65, 67, 68]. Set for formal legislative proposal in Q2 2026 and targeted for adoption by 2027, the Quantum Act aims to unify standardizations, pool transnational resources, and financially incentivize the manufacturing of sovereign quantum hardware [cite: 59, 65, 67, 68]. For heavily regulated sectors like European banking, the Act intersects critically with mandates like the Digital Operational Resilience Act (DORA) and the Network and Information Security 2 (NIS2) Directive, enforcing a mandatory transition to post-quantum cryptographic (PQC) standards to mitigate "harvest now, decrypt later" cyber threats [cite: 66, 67].

| Region | Primary Funding Model | Strategic Imperatives & Market Position | Investment Scale |
| :--- | :--- | :--- | :--- |
| **United States** | Private VC, Technology Conglomerates | Dominates hardware patents, enterprise commercialization, and venture funding. Focus on gate-based FTQC. | Capture of ~50% of global private capital; ~$6B Federal funding [cite: 61, 63, 64]. |
| **China** | State-directed Public Funding | Focus on sovereign supply chains and unhackable quantum communication networks. Closed ecosystem. | ~$15B state funding; 1 Trillion RMB deep-tech guidance fund [cite: 58, 62, 63]. |
| **Europe** | Supranational Public Grants | High academic output, low private commercialization. Focus on standardizing the fragmented market via the Quantum Act. | ~$11B+ public investment; captures only 5% of global private capital [cite: 59, 60, 63, 64]. |

## Conclusion

The assertion that Quantum Machine Learning will rapidly eclipse classical artificial intelligence across all domains is mostly hype, heavily influenced by an early theoretical reliance on the unrealistic assumption of Quantum RAM. The rigorous mathematical process of dequantization has conclusively proven that for standard, classical data represented in dense, low-rank matrices, randomized classical algorithms can effectively match the exponential speedups originally claimed by quantum theory [cite: 6, 15, 18]. Furthermore, integrating QML into massive, probabilistically driven models like classical Large Language Models remains fundamentally obstructed by hardware memory limits, parameter disparities, and the classical data loading bottleneck [cite: 16, 23, 48].

However, the field of Quantum Machine Learning is not entirely illusory; it offers profound, mathematically proven acceleration in highly specific, structured domains. For quantum-native data, QML algorithms analyze exponentially large state spaces directly, bypassing the Holevo bound and reconstructing models exponentially faster than any conceivable classical counterpart [cite: 7, 8]. Additionally, novel streaming architectures like Quantum Oracle Sketching (2026) have established that exponential space advantages are attainable for classical data processing using as few as 60 logical qubits, circumventing the need for QRAM altogether [cite: 9, 24].

The realization of these theoretical frameworks now awaits the engineering transition from the NISQ era to Fault-Tolerant Quantum Computing. With major hardware providers slated to deliver the first large-scale, error-corrected quantum computers by 2029 [cite: 53, 56], QML is poised to shift from an academic exercise in heuristic approximation to a specialized, commercially devastating tool for physical simulation, cryptography, and high-dimensional sparse data analysis.

## Sources
1. [Machine learning for practical quantum error mitigation](https://research.ibm.com/publications/machine-learning-for-practical-quantum-error-mitigation)
2. [A quantum-inspired classical algorithm for recommendation systems](https://arxiv.org/abs/1807.04271)
3. [Quantum Advantage in Learning from Experiments](https://research.google/blog/quantum-advantage-in-learning-from-experiments/)
4. [What is dequantization in Quantum Machine Learning?](https://medium.com/@deltorobarba/what-is-dequantization-in-quantum-machine-learning-a3b4d5af0f0f)
5. [Dequantization of Recommendation Systems (Thesis)](https://liu.diva-portal.org/smash/get/diva2:1665668/FULLTEXT01.pdf)
6. [Quantum Machine Learning: A Comprehensive Review](https://arxiv.org/pdf/2401.11351)
7. [Quantum Machine Learning in Applied Domains](https://pmc.ncbi.nlm.nih.gov/articles/PMC12053761/)
8. [Quantum Machine Learning Algorithms](https://www.articsledge.com/post/quantum-machine-learning-qml-algorithms)
9. [Time in China](https://www.google.com/search?q=time+in+China)
10. [The European Quantum Act](https://www.european-quantum-act.com/)
11. [EU to propose Quantum Act](https://www.moodys.com/web/en/us/insights/regulatory-news/eu-to-propose-quantum-act-to-govern-and-scale-quantum-capabiliti.html)
12. [Quantum Policies - European Commission](https://digital-strategy.ec.europa.eu/en/policies/quantum)
13. [Europe drafts Quantum Act to boost tech leadership](https://www.electronicsforyou.biz/industry-buzz/europe-drafts-quantum-act-to-boost-tech-leadership/)
14. [DIGITALEUROPE Position on Quantum Act](https://cdn.digitaleurope.org/uploads/2025/12/FINAL-position-Quantum-Act-CfE.pdf)
15. [QML: Navigating the NISQ Era](https://medium.com/@nirvana.elahi/quantum-machine-learning-qml-navigating-the-nisq-era-for-exponential-ai-advantage-bacb0c5fe737)
16. [Benchmark study comparing variational quantum algorithms](https://arxiv.org/html/2504.12416v1)
17. [Quantum and Classical Machine Learning Performance](https://scholar.smu.edu/cgi/viewcontent.cgi?article=1047&context=datasciencereview)
18. [QML for Climate Data](https://pmc.ncbi.nlm.nih.gov/articles/PMC12321982/)
19. [Quantum vs Classical Computing Limits](https://publications.eai.eu/index.php/IoT/article/view/11029)
20. [Role of Quantum Computing in Future LLMs](https://www.techaheadcorp.com/blog/the-role-of-quantum-computing-in-future-llms/)
21. [Multiverse Computing Cuts Perplexity in LLMs](https://quantumzeitgeist.com/multiverse-computing-cuts-perplexity-llm/)
22. [Quantum Computing's Biggest Challenge](https://medium.com/muserscribe/quantum-computings-biggest-challenge-a4c1089f3bdd)
23. [CEOs Break Down Challenges in Quantum](https://www.louisbouchard.ai/14-ceos-break-down-the-challenges-in-quantum/)
24. [Quantum Computers Will Make AI Better](https://www.quantinuum.com/blog/quantum-computers-will-make-ai-better)
25. [Role of Quantum Computing in Future LLMs (Update)](https://www.techaheadcorp.com/blog/the-role-of-quantum-computing-in-future-llms/)
26. [Quantum Timelines](https://introtoquantum.org/essentials/timelines/)
27. [Quantum Readiness Roadmap](https://www.fdd.org/wp-content/uploads/2024/09/fdd-memo-quantum-readiness-roadmap.pdf)
28. [Alice & Bob Roadmap](https://alice-bob.com/wp-content/uploads/2024/12/Alice-Bob-Roadmap-Long-Version.pdf)
29. [Quantum Scalability and Solution Quality](https://arxiv.org/html/2601.08578v1)
30. [European Quantum Act Current Status](https://www.european-quantum-act.com/)
31. [EU Quantum Act Strategy](https://www.moodys.com/web/en/us/insights/regulatory-news/eu-to-propose-quantum-act-to-govern-and-scale-quantum-capabiliti.html)
32. [Turning Europe's Strategy to Growth](https://www.telefonica.com/en/communication-room/blog/quantum-act-turning-europes-strategy-growth-security/)
33. [DIGITALEUROPE Quantum Support](https://cdn.digitaleurope.org/uploads/2025/12/FINAL-position-Quantum-Act-CfE.pdf)
34. [Europe's Quantum Challenge](https://thinkeuropa.dk/brief/2026-03-europes-quantum-challenge-the-eu-must-ensure-rapid-commercialization-without)
35. [Classical Data Limits Quantum Frontiers](https://quantumzeitgeist.com/classical-data-limits-quantumfrontiers-quantum/)
36. [Exponential quantum advantage in processing massive classical data](https://quantumcomputingreport.com/exponential-quantum-advantage-in-processing-massive-classical-data/)
37. [Data Complexity: Classical vs QML](https://arxiv.org/abs/2509.16410)
38. [Classical versus Quantum Machine Learning](https://medium.com/@gwrx2005/classical-versus-quantum-machine-learning-a-comparative-study-1e65ced52a42)
39. [Quantum Data Loading](https://www.bluequbit.io/blog/quantum-data-loading)
40. [Quantum-Enhanced Machine Learning: QSVM & QPCA](https://medium.com/quantum-computing-and-ai-ml/quantum-enhanced-machine-learning-qsvm-quantum-pca-17dec03315c1)
41. [PCA vs LDA in QML](https://www.mdpi.com/1099-4300/24/11/1656)
42. [QML for Binary Classification](https://eprints.bournemouth.ac.uk/40696/1/software-03-00024.pdf)
43. [Comparative Analysis of a Quantum SVM](https://www.semanticscholar.org/paper/Comparative-Analysis-of-a-Quantum-SVM-With-an-SVMs-Hidalgo/dc1d4843681e074596321ccf1b3ceb0767841a39)
44. [QC Ware Machine Learning](https://www.qcware.com/categories/machine-learning)
45. [Barren plateaus in quantum neural network training landscapes](https://scholar.google.com/citations?user=yVy2ZIwAAAAJ&hl=en)
46. [Simons Institute: McClean Slides](https://simons.berkeley.edu/sites/default/files/docs/17775/slides-mccleanjarrod.pdf)
47. [The Development of Quantum Machine Learning](https://www.researchgate.net/publication/358178757_The_Development_of_Quantum_Machine_Learning)
48. [Introduction to QML (DTIC)](https://apps.dtic.mil/sti/trecms/pdf/AD1212942.pdf)
49. [Machine learning's quantum leap](https://www.advancedsciencenews.com/machine-learnings-quantum-leap-into-superconductors-and-quantum-computers/)
50. [New algorithm unlocks power of QML](https://www.advancedsciencenews.com/new-algorithm-unlocks-the-power-of-quantum-machine-learning/)
51. [Q-NEXT Publications](https://q-next.org/publications/)
52. [Multi-target Quantum Compilation](https://www.sciencedaily.com/releases/2024/12/241210115620.htm)
53. [Review of Concepts in QML](https://arxiv.org/pdf/2401.11351)
54. [Google Quantum AI Research](https://quantumai.google/research)
55. [Quantum Computing and LLMs](https://dev.to/aaronsteers/quantum-computing-and-llms-match-made-in-heaven-35fp)
56. [Convergence of Frontiers: QC and LLMs](https://medium.com/@ajit1-patil/quantum-computing-and-large-language-models-a-convergence-of-frontiers-545ad5b81101)
57. [LLM generation of Quantum Circuits](https://arxiv.org/abs/2506.18125)
58. [Quantum Computing for LLMs Survey](https://www.techrxiv.org/doi/10.36227/techrxiv.177004376.69666864)
59. [Debunking LLM Quantum Hype](https://www.youtube.com/watch?v=nfMZ0zJ61DA)
60. [The Year of Quantum 2025](https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/the-year-of-quantum-from-concept-to-reality-in-2025)
61. [Quantum Computing Industry Trends 2025](https://www.spinquanta.com/news-detail/quantum-computing-industry-trends-2025-breakthrough-milestones-commercial-transition)
62. [Europe's Quantum Leap](https://cepa.org/article/europes-quantum-leap-challenges-us-dominance/)
63. [Global Quantum Ecosystems](https://luxinnovation.lu/resources/quantum-ecosystems)
64. [Worldwide Quantum Initiatives](https://www.qureca.com/quantum-initiatives-worldwide/)
65. [The Quantum Data Loading Bottleneck](https://medium.com/data-and-beyond/how-data-movement-not-just-qubits-and-algorithms-will-determine-who-actually-wins-in-quantum-ai-83d602ebbebc)
66. [Quantum Data Augmentation](https://medium.com/@mashapotatoes/quantum-data-augmentation-resolving-bottlenecks-in-quantum-machine-learning-12bf9acfc037)
67. [Data Loading Constraints (FRQI)](https://aimultimedialab.ro/ai4mm-workshop-2/4_Alexandru_Ionita_AIM_Workshop_2025.pdf)
68. [Comparing Data Loading Methods](https://arxiv.org/abs/2407.15745)
69. [Hybrid Workflows in Cybersecurity](https://ceur-ws.org/Vol-4124/paper46.pdf)
70. [NISQ Utility vs FTQC Timeline](https://introtoquantum.org/essentials/timelines/)
71. [Navigating the NISQ Era Outlook](https://medium.com/@nirvana.elahi/quantum-machine-learning-qml-navigating-the-nisq-era-for-exponential-ai-advantage-bacb0c5fe737)
72. [Quantum Computing Timelines 2025](https://brianlenahan.substack.com/p/quantum-computing-timelines-2025)
73. [NISQ vs FTQC Concepts](https://arxiv.org/html/2401.11351v1)
74. [Quantum Market Cap Roadmap](https://quantummarketcap.com/roadmap)
75. [Quantum Machine Learning Explained](https://www.spinquanta.com/news-detail/quantum-machine-learning-explained-from-theory-to-use)
76. [Classical vs QML Features](https://medium.com/@gwrx2005/classical-versus-quantum-machine-learning-a-comparative-study-1e65ced52a42)
77. [Impact of QC on ML](https://www.ionq.com/blog/the-impact-of-quantum-computing-on-machine-learning)
78. [Enhanced Cybersecurity with QML](https://hellofuture.orange.com/en/enhanced-cybersecurity-better-ai-models-the-tantalizing-potential-of-qml/)
79. [Transformative Benefits of QML](https://www.ingenii.io/qml-blog/transformative-benefits-of-quantum-machine-learning)
80. [HHL Algorithm Details](https://www.quandela.com/resources/quantum-computing-glossary/hhl-algorithm-quantum-linear-systems/)
81. [Quantum Linear Algebra HHL](https://medium.com/@adnanmasood/quantum-sundays-12-hhl-log-time-linear-algebra-when-quantum-logs-beat-classical-cubes-79348dc66b8f)
82. [HHL Algorithm (Wikipedia)](https://en.wikipedia.org/wiki/HHL_algorithm)
83. [QML Overview (Aaronson)](https://www.scottaaronson.com/papers/qml.pdf)
84. [HHL Speedup Scenarios](https://www.preprints.org/manuscript/201705.0128/v1/download)
85. [Quantum Machine Learning (Wikipedia)](https://en.wikipedia.org/wiki/Quantum_machine_learning)
86. [What Makes QML Quantum?](https://towardsdatascience.com/what-makes-quantum-machine-learning-quantum/)
87. [QML A Quick Overview](https://medium.com/colibritd-quantum/quantum-machine-learning-a-quick-overview-f7437b368d43)
88. [Quantum Machine Learning Intro](https://www.youtube.com/watch?v=A5QTukT1VS8)
89. [Fundamental QML Algorithms](https://pmc.ncbi.nlm.nih.gov/articles/PMC9955545/)
90. [ML for Quantum Error Mitigation](https://arxiv.org/abs/2309.17368)
91. [Early FTQC Error Correction](https://www.jst.go.jp/moonshot/en/program/goal6/appeal/files/61_koashi_ap03.pdf)
92. [Data-driven Error Mitigation](https://quantum-journal.org/papers/q-2023-06-06-1034/)
93. [Top Quantum Papers 2024](https://pennylane.ai/blog/2025/01/top_quantum_algorithms_papers_fall_2024)
94. [Novel Approach to Quantum Error Correction](https://ml4q.de/2024/03/novel-approach-to-quantum-error-correction-results-published-in-prx-quantum/)
95. [Dequantized QML Algorithms Thesis](https://liu.diva-portal.org/smash/get/diva2:1665668/FULLTEXT01.pdf)
96. [Low-Rank Dequantization Framework](https://arxiv.org/html/2304.04932v2)
97. [Quantum-Inspired Sampling (Tang)](https://ewintang.com/assets/2019-01-28-an-overview-of-quantum-inspired-sampling.pdf)
98. [Understanding Quantum Advantage](https://www.researchgate.net/publication/363273838_Dequantizing_algorithms_to_understand_quantum_advantage_in_machine_learning)
99. [What is Dequantization in QML?](https://medium.com/@deltorobarba/what-is-dequantization-in-quantum-machine-learning-a3b4d5af0f0f)
100. [Exponential Quantum Advantage Processing Classical Data](https://quantumcomputingreport.com/exponential-quantum-advantage-in-processing-massive-classical-data/)
101. [Quantum Oracle Sketching Review](https://thomasehmer1.substack.com/p/quantum-oracle-sketching-exponential-advantage)
102. [Quantum Oracle Sketching (ArXiv)](https://arxiv.org/pdf/2604.07639)
103. [Unleashing Advantage of Quantum AI](https://quantumfrontiers.com/2026/04/09/unleashing-the-advantage-of-quantum-ai/)
104. [Massive Classical Data QML ArXiv](https://arxiv.org/abs/2604.07639)
105. [IBM Quantum Roadmap 2024 Update](https://www.ibm.com/quantum/assets/IBM_Quantum_Developmen_&_Innovation_Roadmap_Explainer_2024-Update.pdf)
106. [IBM Quantum Roadmaps](https://www.ibm.com/roadmaps/quantum/)
107. [IBM Quantum Starling 2029](https://postquantum.com/industry-news/ibm-quantum-roadmap-2029/)
108. [IBM Roadmap Video](https://www.youtube.com/watch?v=d5aIx79OTps)
109. [IBM Quantum 2033 Roadmap](https://www.ibm.com/roadmaps/quantum/2033%2B/)

**Sources:**
1. [wikipedia.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQG2RMcknoKIVuE8_8S8XGBSwzM5OC3OnW7abEdiqbiaqpr5Ckyhr7Ivs2Dk4kTjQiwhjpugdSuo45KpCw4epp82exfqTykMpuaLppuDvIgIrjc6Xnt09B3U7_F7hNx05WbprRmjJSx7qlx3Pr0=)
2. [towardsdatascience.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHeJcg0Ax8x8hfBN1OqpNjq_vdAlfCA1DJ6OmTqPlmJfN6-XkaF4GE6nzp3C3DI7qF4_MO-EDQkl-epiK7m54gQpHUcdoEofqeeicYgTac2ExIl0P5Co0E4_FN0hsnrlholyMwZavUEEqMH1tnJdWvsEAXokrDfFt0fBXxWj2ShVHw=)
3. [articsledge.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQH1is_bXE6Oi2EbDbYXA77lYibv2NWcapYA69FFZz9gGXIgF5F5FYFWUxHrAzyDH61BmmAfe1Ygsfs8CnBRvY_3JI-INcIw3G-kExmy2wYqzzQwUJvbr53VsWDRpbzg_SkieMBICEjr1Cjzgs5z0cJEcDQWAhXyL4zn_u4FvPk=)
4. [medium.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFeILxQNfN_GDS0hal9BiodLNC-rpIF67-iPZ7sMVpVpVkYGXe9J_okwV9aax0eUnrIgzuVfRP_jHyCwBRnjj54mZPg0z05um_RjGBK4UhN8fOBRbetMRW3SlqDUnJLmCuhApsQvMBNU7k8gYOwJ6x4--GK2XJF78AWi9jyH72bHqoQ_JSZsa0D1TQ2TJovzctVSXoi37a2M4jBrVbdNGepz3Kk0kCvJkWWDLPP2cjMkdp_NbgI8IE=)
5. [medium.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGK2b8WQQt6y_iN0fscdMDuSPwu8cMjexKfM59uCU0VjAVG3dbURHj8x5kg6K2nO_Ou3UAUbHeXN3kt7S7d-heIB_6QPz6sekmyrk7aP8AX32TFE3moTGnU9we6ufnhCBL-ZvLxeL-QTUtUkW_paKyFIj0nXy-Y_ctAWySLPgCYWfWDE66IMuvx9TBxoJ4veqo4DqcDKTx1gs5bUkTr)
6. [Link](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHZBL4HLRi5X2HEYya3_MhHas978i8yyzFXa-BJtHkYhPFAC3z7_pHwAsoSpVgBdq7UHTtnqlnJU5Z3M3IiygUs3irevY5dvxmAFhk3cuQX5BHd94uwrg==)
7. [Link](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHujcplQkihY8dYHW4dlePgskJMIGXLPkrc9KFR-Bxj1MEcD3X9eCYmo7dceLmcaiGoZJbhz4k4AzEfn_N2-nidL82ZzpYph0oQZTbCi2ZiHFvfYD843t5jhI3imBgK0HKGfc4RRYkQuNVNWMZbvQH0yaSeY2fFCfc19OsjafvPxbDZ)
8. [quantumzeitgeist.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF-sSrt_joRnSfd2a0QP-9AlqTMng9eBMOJYkU_iT7I4sorf-f65ZV273906jvZPqfIwPbtjPllmAbJ-YPjCZ16u-KYNwzk-seT6cIoLWwu9DCjFp-JLHAH0rXNjeEBRd6b2jGCjR2ceHIxHMhddD91FZrOvyTYRvHzdOjF4cnWBGp8)
9. [quantumcomputingreport.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFk7VIlBlwtOINxWGEs29DwVgpl3XJctBwLDLKa_jPBWiSUbXCn3uyXu-sj3KkgRWRXcz3yqMTMsi7pA2UuxSnII7vyFCSQh1CD5y5RRIiU56_f_QOlrkk1QoyP0dnKLOHyCA-NX-t90o_XySxnP0KXf3tGBT-sm_85wpJa9LzkCZt11bAf_obniTwLE2HPiXqlF46O6NrPV550i_0=)
10. [smu.edu](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHf1SizYaBgoFiVy0KBcxj_7n1Nft1QieNoLydp3lALtZwAC2VeNx0JPRU-FY0_oBpBI0HkGzpWvxAXdRRSB7T4_Et4LM8MgZoH5V_tdnDZM2ZDOHuSxdiJ9eDrypeyh1rZplpxX8F6qfBv5unA89ej0o6gaW_kKCCxGOEoH6pCso3vKxoJOjB4)
11. [spinquanta.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQELj87YyE8Qefy0sPCcYBCINjyfNrXsng84z1RHDVd3Qy7DjQ6d-GyzwlSiwO1Ds5ad9qUPhg6K_0t5VdPZjHsDn_KUEgKzf691V8i_N3gHiRyvSgewwGW3kwbIBYoSbZUoZQmwKORCB4dN3egFYfTuA6t0F-VcwddV2MsG9TY47KYkHzyagiRxMWypxlW8_wHeWA==)
12. [medium.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEXvRwrkK8hErLgv2TyaRGkfm6sEc78aLFCz-rBCZd2diHb7l3mOEgdI-wlaV3bRHYrYbMmm_tLcoEBGlA4w8_UaOxw9Vn-hC5HRe0yIQG8nFdM2x_WFb4zqZpZ--PtEkzeQJpoov7NmoJ7AXlU8GisTLIEbsT2CNQf7hRp6jvjvFsnJyhQnMaI3tyli15Vi3P1)
13. [nih.gov](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHLY4vebaY-_ks2aMRuFzTXRwetQTuoNk8hk9vK4XCdSQgzkPHLB_OPkncmXpFlu1FVQE4Vdi-JzO0G5vCDGZzUeZwmOdHY1LDWEboSHb5l4jAcWkdvocUSVc9hPvYiAzhG61ZuvCVy)
14. [diva-portal.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEtuwnZMGm8WBKXLFnEdqLFPcKHrN6J5Xt7YQwxYpQxsqa0YWvJk0mgYAiVMFMVOQsO1P3HRyFeqGYmyY_nfWiyn-4J9tW83VFLYodEYbcSjezi3KRBRuaawmqagRG_oAXDLONyFkTbR0FmOQ2MRUedHH0rpe6NLvw=)
15. [medium.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHLgmfk4qLb-PrxoKUIRwa6xAYENRutomrmt1iRiM2YJGxRFxyDB73GX4vYkSmUCmI-nEVBwmqQlLpA6KcD5_m-Vx3H_tZnIEflNMnTj1_KV3jSqGLJTesHiTUz2zDW6ARNHuRbIV7e0si-YHisvvxc4959n-FYLmu_xmaiZe-_11Zv2wLNlex_CK7NemrvBuoLcy8_AJA=)
16. [medium.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHi8v_rYge6qS2yEeYZ5SK3AIiQrLlfJasG59eag7yvZvA09XV2p2geJQdbylRVvz-4yHPKOPdCdlB01NZBCvPYu2W57Jb0Qqbwok8qh5iYQSA0If_9Sc-TGwcqAcQDnX392a35MxQwLzm8nfolTEa4yPyq714LZHVaOqwP_okmq7zqhrH3gSoCfNkMfqUFuABDehdnfqTWAqC4HBScCZPFqVRjgojTfxvZ2rjHnaQxs0-2NqJXM3T78QwoNgFpZ9jHWETMikg=)
17. [scottaaronson.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGW7F_5G6V9XvqzamniGr1nFKPXKtCRuqkA2CFBgiX1m2ePGdZKQBWs_WN3-ft19A1G383x3tbYlDI6ikeLY4Xz-KG5C6x6dbOXt-vTiwXoqEFSmVdBqpwtRxPpLg-1fKmDxA==)
18. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFEMHDJgHZo76fHtS0rmk3GWfkT7YxTksGsS58uz3E3sNvJsHjTQjy8ufsUKMCDxCwmT4bdN8Q3WbHxKorljPmav7AiuexUNJsNK00kFVsOkJv62IoF191EwA==)
19. [ewintang.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGMoCA1sTJ1cvqekExxR_9HzEXHbEtj0ry2eBTtBlhpbZRV07lh5R52eyaVZm2t3-y8iDV-g5xPu-NawYng6IGnZ3Rwq1xtwdv7wP2_PPzhgz13ag6ReECjdCuP4lT28xqqOFJlAnwVpU2j918km2sAaWKk2WZkZNAuCGde8gOb6Zi83nponnE0Iw==)
20. [researchgate.net](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEkimX7LqdKycEJBroeT704K_huuASoEc5GLkD8kwFSxlWNx-g4xLWGzewlrjrzH8jI6Ns-pYVXI9Xhrk7m5aDp7uKv8hllJj_ZSuIIZJhXJYqu1puO-uFgZF-utPEHRMZlIlM9ikfVxfXdPtQn6ENux8tRt-ZOXm2OliRkw0Yus3ugZLogGFd-623bT5EbrRHtI-SNxmfTkTiPdZolXW9iiJhidUPpeqbOaBPBHSYih6ePAFE=)
21. [wikipedia.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHYykEdylYs0RrqN5q6NmlEzcJpwXMXvumw7bn5H9XsdeaDRXH4G36avY69aTRvttdJJ_Jqw1TI1FLBs77m7urXZzNn8wGshp_El3_UL_cT4hJ_DVYycydy4eO1p6Ju_usR)
22. [bluequbit.io](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGZnL-RQIG0swEOHDWhSACfQJ1qH_feGQUVXNwqu8J1st2EFvFwQX7v1EfdoLGSM4MSh_Mp-SH17Ow9oD3kin37L0_llahNNl4sjZ8cD8Fu6FiXcZMKXvCnLjAI2R9kzO5QMAxMf0htGw==)
23. [medium.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEY0deFq4kc_FwQ4HNpecSRUIifHJmAYeqsQ1vXk5rMlgUGHM6Q8wGgb0gVipvS4TaU21qz2Cc7-tbnXNMddWZ-hh7OTNgf5BbEflE2esU634WOfWjMJKsfV0ndLgPwG48EFMcS1nTc8Ysj9gg5IgvOK4gG4xWfY17uAnq6lD9UsgKRI9HhBWsMyqnansLgktq9s0gdL2MQGZnT6kiuI6Bj-XN4fml1HbS3glwKD1KuWw==)
24. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFJk3yIb0r4v52ht5z5RF5HIfFSBBLCd6RQBnpj4ucJvtEfqxBkjN3ZRZOa9bE0O-erBpyya6GCa7u1Lr21mcqSphFAriKiBFWE-4JgCrFHI8Vg7IlAkA==)
25. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGuen1UrYvCJwGfunzJtbKANlbkSrJ5b-4mr1rsrkeldMxSzusVIQM7WmN4oboBgS5CpMzqsfYJEzVDoLEBj7jILPhmfVxwSRhzOBv-GtxksjDOjcxxvA==)
26. [aimultimedialab.ro](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFKlWM_F4M2KffJdiVvOK8ndTf3-WTFkw3LMHD0tDl-wJoQNsqNPOmQi00S60-Xue_xrg0uUVtCA3GddQgNXibFiyZynRp1Dq2smoQ5Uwxrz1c7t2mKGX9sy_YHGSXNKhj19abuE1-oixnlCF3iuU__1bwsz-H2DOjoQ0joq22UW-G-ugVEIcnqkZE=)
27. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGwtWhMqDbYlE62_bh5fBfXzOIf-3DCFxOnVvgzGLUwsk1amGReHbrtQ-8WHDphoOGWVrniRgg8fRMeSsyfnQ_8p6162qmSAQ_9HP5XeNH9bLHjn3bVs-UCFA==)
28. [quantummarketcap.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF3QDwf3kpJWF9sF8iFC0rgkST2RDXToHhulAjMdk6dcSD867iDu_kdgMfvHQGS8HHlbXEXnCgyM_jAwJWNPMsMT5gP3MdPgsOSf6HosQuqwMM7Xx4nFDKzb-c=)
29. [medium.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFd-7lTZDzSOjw5gIHxtA2QoKQSWKIJJySNDHKRcqdyiuYT_v-ldoDIV1qWTnBjkP5ZZ6ACAfxQBxsKGEAhyu3mn31fW6LcK6jtzChbq3wiuF3pP4XHZaas321oxcxZ-4Rhy1EiQM4A5P1Q193vtV1v4hHOjwq7_QgZ8yqR7JQ4U8JPhZQUyw==)
30. [louisbouchard.ai](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEQ89YYrBNNsjM5yaEjw6M87SCEV-O-xuEJDI21sSvDmU6NsBhxqP0gOa_3YJRCX9io-jtDU8snOLKmojdTPWnWdSikUbcpYzU5lKL-VyRXAMXwc5qiwQusoLLz1WwdkYgVZ2wIQiDKypMMLWNExO5Mrg2mjZYNJNUUYMlYbdHiMQ==)
31. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHq_AZxVuVaGR8CJIEHDefaeF9bdLZz0GL-NfOM_Ad58gez6Hfzs307EKftuNFXMB5YyuRjv22RO8d51QAa3Q79Tf7nU4zekldKCYnJF0fNFIBRHyqiCw==)
32. [google.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGgL-XSNhXJakt9mFtKfaJEQ0vKgNhmiV-0w-VWT9O-T-1E4j7azQZfzNzFe-xGy6MCBzZ9TPs30IoikVnyzuMpAjbxZoSF2ZFZN-lOwuJFJVWcJYD6Rmg5r7Y0FU6oaeGBv0jQ9BfGk4jhNOQfJsYt6P8=)
33. [berkeley.edu](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF53fwlJPWjDeeu2VGkcqyzcDC-dd01kjQtmaxql1ow4ZvPjwWDgIiGGtr1NbP6cHNkqFjs1UgzohVjyjwUueQzckHDSDVegsPSBE-uNtPAdNQ6LtAnQ0UsOvfLe3qqBcFJetTmJt8TgUcohYsr1Ebbw1qcDX-vBdUyf-JAZJtOgeaoO02l6nyZvg==)
34. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHhw6YYtNXWaL_dJxsZHWc_Y0z4KxyUi2ngtoJ8qdkXaY1G9OuF6QmA4N577RTFjDvtafczTgK4xUpwJi6BuEATd7ENeOw4_s5sHU1WenmPdvRr9-8gxJw36w==)
35. [introtoquantum.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFVcFZL_D5Hp7E8xW-OcyTxDkBiahGZ_WgR__0RDPs5vOa1cVMQgQ-Z19k3AbFlAwhmzkiB3ROXYXJtbEJ09hbFsyLjhQp9fFSssskEi-bX-6a8CSNm8vN2t7X1JZ04RhSIx3BWnjg=)
36. [medium.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHdrJEC0kwE5XyS6Nc-clcJJUObs2l70bRjiPDHA0EAcNMCJuQeSl5I2pZqTNPde_sQtfDyVfSPI4pPYEWvqTEXnoYs0rA6R2POZ3I29bOmzX4qdPYdwcom3-Gn_VuDjihzSzgf0UoK9BYcl8S_QkxbGY5xaeO2ZEPIbbyRrxyyX8_R8NddawRy714H8tjEOafTFeMhDEEFEeLiXwrb3Qmf2f3OdQ0g-Qw4XGCGdxfTIys_2m_2cn4KYg==)
37. [quandela.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHUcLVC2hgnPxTNzpVDWfC7ain49Q8zK8NnO1L2X14tcRC0QgqcvqEph_8T14b_XPNZUygRcrMa4BiOt_aJPecI-WcktpHLn0BLqFMuiEWo_A4jl_8vG5NxJ_XhPWXeayiNp2ld68ayMJH4yGxR43tA-rHNidUokZpIdW4xdHd2yLyIzpd5vJr-xbt9yOS2YPa6qzOfde62SII=)
38. [preprints.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF5aB68EtIvOCklvmIXHpVGX2HUDu9Uzm_mfAVKnXCQzaDP8oIkR4T1IZjK_u4PKqMn2kyNypuoz_0Bfrc6oTAiY1ZAcTu_a9rEK9qcsnpArjWN_yE61JI_JRzS8gzE9FxZruOJwdfVmORDWWp6l5gJitQ=)
39. [bournemouth.ac.uk](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHlg6dq7IQDdiROOm7Tkun4jgR2xlFb-8hIL0szWhE-q2LK9lkb5KYt0AN7I0sCoDII75YsJqH21MPoDDMomqF1b2oZ3HsuXyL6H0egu4REc4Tm_MtJE6P4zMir8Y67NmFA0zRph-I7imzVavP6L4ClAUT-luk=)
40. [quantumfrontiers.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGgSyh5gKIHNk2vlx7eB1pH6UkFEKsjbj-MnVL1FpILl-sBb43bqEP1_BIHVEYQmYANEYPyUxLpQ6VAemWaPQg8LGRCJGmFRL-Kv91TN3-rU3Aamm1sEYjGNUo_kTfjk0y9Pul26X7_BXFN8WuECNEYxAvR-AgMX3_pZ3Ztc7xY-OYdDVsa)
41. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEOJjPCIhsYpYh9v2ZOoylhW2ZXQw-1oEfkcSrTpa1OpUmeDlpFE1Vds-GxFyHPUAS4YI8tuVHb_75JMgxgIay_uTUO6r73PdKbU3qCYG8vUKYHDQ8grQ==)
42. [substack.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGS5unkyWDGmDWc4Z1H1eVfV5lY5iP_92LTcXjtFF3-7CiQzHoQ_eeSbEbDC1bhJEKkAkN4oDXNc5TkRqQj21KLxXOYmeKOeV2N_ArzcxBFntJDitzazoKiGBkJi1JME7sjA8dYJxdo8qSXBfLQe5pWq2cwfgrqJC5d6oVH0eWQydILWXRIeoaM)
43. [quantum-journal.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQE968q5kZYRhOxAdF2sL8rcMoj0qy1frVtF5n7IBzkfC5qx6mDYFZcEwRCOrlE3IIpdjCNbrJzENlCflGXIJCUjvo-to-23P7ney6g3faThcYkF3BVJXPGwvZs1bL-tVhvhobw2VDpbVBh8Yw==)
44. [Link](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQETg05g1jS_g9D0roQTTLhj0RJ5JA4qIVYjXETgew-47xUL9QnIp3d58-6y9oNe-M358RRG3JLOfgCW2biOj8WTN0heVzGT0j2abJeOuPA0QPfGIPLGSDuLrAstfmR28G7LgO_HBAtJfAaoxtRiwcjDRfUu53W-r70DWfc-vXMN5SVRBUxhhHYQ8jznoS4BMo20b38=)
45. [arxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEqYqBtEFJ3jQKw4znfPNAIqHpD-lzD5MGy5tXp-dRmezS9MbjJkJzKQzw7FHCUTW5_jRMy-L3A8Wd_jy2ksQGcxZU2NVdy-VYKdDhw6XLEB7KhLzhJxg==)
46. [techaheadcorp.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHvtQW0SjRBP4BHNmS9Kud3-UGeoE2QqWXJy20RbeuO2p6Jybls7oTMe3gUQnqJK0uYOPONzR0NhpEy8IyMN9llQHM2iAT1yhKq8r6wOIFlyWkL6tHwDrSCWoXuDsiWzDUTWmiZm0zzLua_2BFtvJ2bbMUHezCTJCNZ_6CqP4y6GKNNmt4gvA==)
47. [quantinuum.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFxKcwbmVfpTjHX2JqA8VItwdHpTb1IbuL__BbiT0Y62RyEOhUO6Dpmx5gSibVZRyBJbnlk35ueGs8wCX7bgcHa4WcN1vi1PehvIyuotS-qxQ0OvNsL_q1NGPDuJEeLtYgGy4z85Tf4tvql-_Tr6DsYx1ppp6zcpBNNFlw=)
48. [dev.to](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFqbncW-2oqg89LxDJm320bXH2Kax87fKrxutkqYXkrQbMuTQz0M5Bq8NZuClpG5OLIk8V8aREYv4OmYNaL3FVAtrQGD7wP58zaOVT4IaMXi80x-7hLXF5NwK7LHjrrewM_phliD8WEc98ZsBmwKZPSLCBvUgEcgHWprUw-OyvxLWX75Xys)
49. [medium.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFQkKxDG1sCaCbozgOyK-UX4oEi9uabl9mYyMzqVPZ2PaC4CYCDVJh-l5hjYIzJlePL-QD5GoMZLEhXp6RCBZUDrZqwpWQgzw83Nfa5WYXvGp-H78K0jxp4dmKRt3n0-qGPSeeswOsKQnmS7bKoafIQ1li_I00phf63pumBZP1teIAzRruudVvcgwBE5lo7DFiIL9TBG9zlas8VnkQwpYYqoZ6usmA6FwHR)
50. [quantumzeitgeist.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHiqAeY1vZOoVjf3p8Orl9OK--1JgOspex-nkZdv3UZ9swv7WQDEZy829rIcltjVkxpBwZoa3MJ2-NRDkMpqYlS3d9I3k5B_7KnlMkOHoU84WPSUJ7UaoexSmEl43OWQ2vKgooaRgLIcQd8EIMp8Nk1-T0U19gvTuSCIUKH)
51. [techrxiv.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQETuuKEoGAYXgHXKxZmklgiW1rXmJcSSqyMfRcuUtuWOzsCgF4VJ9XaLrHlIz4VrEHSdxdzqqHuIFUQ1qIfJN7yFcwoSck4uw5nXznbSycinirnohska3pkqVZd8LT8PuWgNtHkgg663b8e1MtP3ekExE96CQGJuQ==)
52. [substack.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFMJHMBEtP8hKX9aG9pFeTmKswa9BP-RwtAD64ObIn-TwRUSoKBNYpCu7v_yVmCy66ohXByacQvQQK6Jw3nBmhNZp74-e6s5MeCZWLUmk6VJ6KmTKfr7XWsBBN8y6YE7Zjq9MEQsO8jeLSNW9mbZ_gFis7zetxbrDKTKw==)
53. [ibm.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFrsYTlqm7nic152HbBX94ZFjkCewhmlk6kmnyqli0SmTHK9kBuOma22bgFXJYpFjBdAXsQ6rhdV2Q0Pg4f8Hvu3NtVZK9qDy632VfOdZd14LRJtX2EMZk76BAeGSgB5262J-XZN3cNdXACtgjacfkPiv3qunL1ITcAInMifQhzxV1din0q99CEcBVwBk5U56VPdNutcd0P-uHgWz43Lg==)
54. [ibm.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFHMkaL3wfX0URpPMXQFOnnIokcQKfxssZVH8gsBKuwRuY1iKvgThbt4lylhbvwb_5ksm92ylscQMEP-WV7RdqF66oY9Vq7ah4Q9Zc5nK0n-JYg_cTHfmap_9d6)
55. [youtube.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHUdz4mZK0ZhV78tiLjC0cVdH8EeySrYwgH8LnB9X6k_LW4vbjT4XrpZr9ibAoAZBYHPpU0FFRh426REt44ZGPZSec-LIm6p8I3dhl-bPshRD-M2mnTyVVXeF3FNzDXqiHf)
56. [postquantum.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEUjV-5WzAhYs8C2E4ye1TL2UQmdo4Mid3KM7sC26tySw-PUmN4JBSPECcif0H70F1lPEiUQscDry37Y0GYhYMxTowZJXaKNuRbdB18Lz3ZZz3BGRCyAHLv_pRYV8vgTg18_9w67IxSofpdfcGYcy1Alxs-Jg4=)
57. [ibm.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEkLJodc9xPMkgIGAjiL3TcwDwMf909lz2bKeHJ1xRKeW-_v7SZFkVBt-za-J7sSSjgNP7DVqcKnmSYsUs6iR4pzx4OMcX2ushMI_qgvHgDQkbb20v3AcOsQWlVCGtuIUoiBSo=)
58. [qureca.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFC7-sexH9csRMQ4Tpf4T6IR71GQjCgu_RjFOhFb1_OEttAbEjfi8icbK6etOggiwVDLnZPoRGpPRLWH1ZQk1WLM433YHA_3LVw1ffMZR-8AZ74BHicOvXrR7i8btVbMhBwv3W6HhNreL6K_Q==)
59. [electronicsforyou.biz](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEO5iIO0y8WV2bSyef4WDSM84rwmdaxxjE92xVHetS8UbAFFioTnQoqRQw_LTu9blFnm6dDHi8ejcm7EWZ_1v8yc4N_IpU52dA4PJcdpCc-6_k0DYe0Ect8y7N5vT06nR1VEgVaYTMHViZjhE1dMD2EBeomB2u342Vwk4C1qmshl85Vx5ChESmZZtZaYz7qdsETQwyqOpOeUrc=)
60. [digitaleurope.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGhN6JCaUtwtmI9PVA80ZFSax_RIDwYwBukke4SgeScGO-4gfWIuluhp-siTjhnZG-f44gCosi7NuhX_LZigSsaGSwv65qGPsh0SvY2rbtc6k_wVWqwgzjfhoseogAZwVvBoUTwoSC13X164qdG0CJxCijwr0Qndkwj8DRWga5vAib3faMzIg==)
61. [mckinsey.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEbknQKwLaLCWFMHY0FEf_E-fmHhsC_ZgRQJ8qj4LQ5IGiqnbiAm3_r1Z8f_yojMxZK6PFKm1pHlfWIlWbtjLdIkkrxE8PBJZCNPAlv0gtNFYyhy8DchEfxpeGm_eBAlRdIpqAARyNGQ81ahvjj4pRrELk9xX0duaMaOAdpi0AjYIY6UW5peLXayFm9QmZKOpwzg8Z9DLd73YZl3cKy2u_pzTcfTnGTITY=)
62. [spinquanta.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFNXqLv-LaJ6Ru2PBTsMwvaLjOpFodbJe8suRwWvHf98ViNk7bLaUhZeIApz98nw6PYWEwV9s-Lloo3anoVdXE82lDdc4xoZrqEs010ZhxLRyEIUL3C4L520THZnhRH4Plpsh-c6VyxVgK5Ap1Soy7Euj-XgqMXeIMtKeMpgNzgBTuUNH_CYMjfY69MAVSUOwvCXNErJ-9jekFdjR1BpYX5kq0e4Jj6dITNKiYMl4oJ114=)
63. [cepa.org](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFW1sYbOaSeWHKevILiC7CamXrrvSojuOTefqk4cASIijnwxxjrMWWwDEW-KQyTbIr7Mzyrcz-4b1sacZhTdBewnpc9Z_XpN3SSX_hxNxIIOS5psJcw0ZOfGxm9tatZmFc2vHl_i2XsxeU3MZvFiLkYJnOb44eFuXlh-DWM)
64. [luxinnovation.lu](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFbVJeMdyxeao2_fafYmtRIAKcutYm56qBcny8cCBBzEQ6SzDKi22Rbp4h5WAI3EgSlCRNXXufYQY1U-9_d8dFcN8ictP_-EN8o7M-zjerWkmNRQu1z9pcfniro3X0VTGFzzeGBrePPzBMLTg==)
65. [european-quantum-act.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFL2l2XYBNNuV1RIhWlB762CY8a1CvqcRYsC5iwq8gMyaAdsWJIfg5ahqcFXxk3JpfO1YxbfRt0kJS5mXK19yGjRXqYQ7vL8z3DMlHojKDdyKPD8esVc4wp0h2E)
66. [telefonica.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEf9wX6fZVtCeKsGYhkGXplqEd0vZM6293MYMoxklM-Jq4PHbHk2YUp-5JsQCPEFuhLTuFBFn2pN6AhIOLSfo2y0kASaZb_yzV8nCKjh6z9Cc428iMj6XFg6KUrkLg9hPQf5IaERFdCiNkTaj8P8EauBmmdjfQii3LV0CVT4jyZRUfHz9bz0PJmitwJBRklT_To1edK4RR-PuthSobD6z1xow==)
67. [moodys.com](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFOMiL-Qz3-YXOhAoVBxauiaf8OuUzbYWPcnoxlsiY9elGAz72yj29GiZClZ2rH2ybjyoFG6fGYIu9m-tcySEilZQg2S-B7aDeB9nqHJs5dcqOE3aRp8LJ7wo66rIsW1GGVgPRQae_56RPp-JJV18HB921sXudFFOFlBG9tk6SC__sxXh6I26yVNJ6VZAd_D-2aQe04coNZfvsFMK9cDjAkkrGMwQxJ1GpKaZIBNOL3lied0Nfx)
68. [thinkeuropa.dk](https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEY_7fPYDvaXaVu1X1xdNM4aeNWdn3rPqL1XKRu09-V2V0RUZcqSbEkPnw_5TkIN2VPL0mi1qPJ6vPTj5sLSgu__T2VtKEQ9ItzAWeV5qVcNLK672w1SF611WwxN4RFNHUeyhpiVeahql9u-uBWxgg6FZavziZPW7fYGFwajShJefh9iyKeqbf9F6m-pHT9nI4L_iEkaXJJgSN6AwqjMD2gILoqejSJXA==)