Google Cloud and AWS Backlogs Reach $561 Billion and $578 Billion

Several articles point to continued demand for AI infrastructure, while highlighting different beneficiaries and risks. Alphabet reported third-quarter revenue of $127.6 billion, up 25% year over year, and Google Cloud revenue of $30.7 billion, up 103%. Its cloud backlog reached $561 billion, up 262%, with $308.55 billion expected to be recognized over the next 24 months. Amazon posted $201.9 billion in quarterly net sales, up 12%; AWS revenue rose 39% to $46 billion, its fastest growth in 18 quarters, and its backlog stood at $578 billion. The articles describe both companies, along with memory-chip maker Micron, as potential AI-related stock bargains. Micron’s non-GAAP earnings per share reached $33.42, up 1,003%; one article notes that some forecasts see the memory-chip shortage lasting until 2028 or 2029.

China’s Tencent is ramping up AI investment but has yet to show a product as compelling as offerings from ByteDance and Alibaba. Its capital spending jumped sixfold as it bought chips for foundational models and AI agents. Third-quarter profit rose only 0.8% year over year, and revenue growth slowed, with the increased investment weighing on results.

Workplace use of AI tools is often happening beyond company accounts. CultureAI’s analysis of nearly 1.7 million interactions found that 19% used enterprise accounts, 47% personal, free or unauthenticated accounts, and 34% applications with an unknown license type. The study identified 252,480 instances of sensitive information shared across 100 categories, including HR records, company strategy, health information and legal details. In organizations that restricted AI applications, personal-account use increased by about 50%.

Research on language-model pruning finds that importance scores can be fragile: scoring final-answer tokens rather than reasoning tokens reduced accuracy by 27.6 and 32.2 points on two models, despite little change in perplexity. Individual neurons also proved too small a unit for reliable scoring; groups of 100 produced more consistent results. Separately, researchers introduced ARISE, which helps enterprise agents learn from failures by saving and testing corrective rules without changing model weights, tools or workflows. Across 14 workflow areas and five models, ARISE with Claude Opus 4.6 improved results by 10.9 percentage points on SOP-Bench and 5.8 points on τ²-bench, outperforming GEPA on both while using fewer model calls and output tokens at similar task cost.

Companies are also building AI into software and partner operations. Exabeam expanded Sherpa into a guided partner system covering training, sales coaching, competitive guidance, campaigns, deal registration and market development funds; it uses partner activity and opportunities to suggest next steps, with deeper pipeline and deal-risk support planned. Factory expanded its Microsoft partnership so customers can run Factory Droids in Azure, using existing code, identity and governance systems. The Droids can investigate issues, produce and test fixes, open pull requests for human approval, review GitHub pull requests and audit codebases. Microsoft Execution Containers restrict file and network access, with policies managed through Intune. Factory is listed in Microsoft Marketplace, and eligible purchases count toward Azure Consumption Commitments.

Investment and market coverage underline the scale of AI expectations. On November 19, 2026, KraneShares CIO Brendan Ahern said the firm had an opportunity to invest in Anthropic for its U.S.-listed AGIX AI ETF in February 2025. Anthropic’s annualized run-rate revenue rose from about $1 billion in early 2025 to $47 billion in May 2026. Morgan Stanley estimates nearly $3 trillion in AI infrastructure investment will flow through the global economy by 2028, with more than 80% still ahead; McKinsey estimates generative AI could add $17.1 trillion to $25.6 trillion to the global economy annually. European equities also rose amid AI optimism: the Stoxx 600 gained 0.4% to 434.4, Germany’s DAX 30 added 0.5% to 15,444 and the FTSE 100 rose 0.3% to 7,799. The Stoxx 600 technology index led sectors, up 1.4% to 1,444.

Infrastructure access is another focus. vCluster Platform 5.0 lets customers request and configure bare-metal GPU capacity through self-service, choosing from provider-approved machine types, images and cluster templates. Providers retain ownership and set capacity limits; vMetal manages physical servers, while tenant boundaries and quotas restrict customers to assigned resources. Users can build setups such as Slurm or Kubernetes clusters. Version 5.0 is not generally available, and providers still need to test whether self-service improves delivery times and reduces support needs.

Key Takeaways

  • Alphabet’s third-quarter revenue rose 25% to $127.6 billion; Google Cloud revenue increased 103% to $30.7 billion, and its backlog reached $561 billion.
  • Amazon’s quarterly net sales were $201.9 billion, while AWS revenue rose 39% to $46 billion, its fastest growth in 18 quarters; AWS backlog was $578 billion.
  • Micron reported non-GAAP earnings per share of $33.42, up 1,003%; some projections put the memory-chip shortage through 2028 or 2029.
  • Tencent’s capital spending jumped sixfold as it invested in chips, foundational models and AI agents; quarterly profit rose just 0.8%, and it has yet to match AI offerings from ByteDance and Alibaba.
  • CultureAI found only 19% of nearly 1.7 million workplace AI interactions used enterprise accounts; 47% used personal, free or unauthenticated accounts, and 252,480 sensitive-data exposures were identified.
  • Pruning research found accuracy depends heavily on scoring reasoning tokens and evaluating importance at larger scales; ARISE improved Claude Opus 4.6 results by 10.9 points on SOP-Bench and 5.8 on τ²-bench.
  • Exabeam expanded Sherpa into a partner guidance system spanning learning, sales, marketing and programme management, with further pipeline and deal-risk capabilities planned.
  • Factory Droids can run within customers’ Azure environments and investigate, fix, test and review code; Microsoft Execution Containers and Intune policies govern their access.
  • Anthropic’s annualized run-rate revenue rose from about $1 billion in early 2025 to $47 billion in May 2026; Morgan Stanley estimates nearly $3 trillion in AI infrastructure investment by 2028.
  • vCluster Platform 5.0 adds self-service bare-metal GPU provisioning with provider-controlled capacity and tenant quotas, but is not yet generally available.

Alphabet, Amazon and Micron offer potential AI stock bargains

The article names Alphabet, Amazon and Micron Technology as possible bargains if the artificial intelligence market keeps growing. Alphabet reported third-quarter revenue of $127.6 billion, up 25% from a year earlier, and Google Cloud sales of $30.7 billion, up 103%. Amazon’s third-quarter net sales reached $201.9 billion, while Amazon Web Services revenue grew 39% to $46 billion. Micron makes memory chips used in AI systems, and the article says its non-GAAP earnings per share rose to $33.42, up 1,003% year over year.

Alphabet, Amazon and Micron lead the article’s AI stock picks

The article highlights Alphabet, Amazon and Micron Technology as possible bargains if AI growth continues. Alphabet’s third-quarter revenue rose 25% to $127.6 billion, while Google Cloud sales climbed 103% to $30.7 billion. Amazon’s net sales increased 12% to $201.9 billion, and AWS revenue rose 39% to $46 billion, its strongest growth in 18 quarters. Micron makes memory chips that are important to AI, and its non-GAAP earnings per share reached $33.42, up 1,003% from the year-earlier period.

Three AI stocks the article calls undervalued before 2027

The article says Alphabet, Amazon and Micron Technology could be bargain buys if the AI market keeps growing. Alphabet reported a $561 billion cloud backlog, up 262% from a year earlier, and expects to recognize $308.55 billion of it over the next 24 months. Amazon’s AWS revenue grew 39% to $46 billion in the third quarter, its fastest rate in 18 quarters, and AWS had a $578 billion backlog. Micron’s non-GAAP earnings per share rose to $33.42, up 1,003%, amid a shortage of memory chips that some projections say could last until 2028 or 2029.

Alphabet, Amazon and Micron draw attention as AI stocks

The article presents Alphabet, Amazon and Micron Technology as potential bargains if artificial intelligence continues to grow. Alphabet’s third-quarter revenue was $127.6 billion, up 25% year over year, and Google Cloud sales rose 103% to $30.7 billion. Amazon’s sales grew 12% to $201.9 billion, while AWS revenue increased 39% to $46 billion. Micron makes memory chips needed for AI, and its non-GAAP earnings per share reached $33.42, a 1,003% increase from the year-earlier period.

Tencent profit barely rises as AI spending surges

Tencent’s third-quarter net profit rose 0.8% from a year earlier, while the company’s increased AI investments weighed on results. The article’s headline says Tencent’s capital spending jumped sixfold to support its artificial intelligence push. It also says revenue growth slowed and profit was nearly flat. Tencent has not yet shown an AI product as compelling as those from domestic rivals ByteDance and Alibaba.

Tencent boosts AI spending as it works to catch rivals

Tencent sharply increased capital spending this year as it bought more chips to support its foundational models and AI agent applications. The Shenzhen-based company has yet to show an AI product as compelling as offerings from ByteDance and Alibaba. Its latest results were mixed as it pushed ahead with its AI plans.

Most workplace AI use happens outside company accounts

CultureAI analyzed nearly 1.7 million AI interactions and found that only 19% took place through enterprise accounts. Workers used personal, free or unauthenticated accounts for 47% of interactions, while 34% used applications with an unknown license type. The study found 252,480 instances of sensitive data shared across 100 categories, including HR records, company strategy, health information and legal details. In organizations that restricted AI applications, personal account use rose by about 50%.

Study finds language model pruning depends on fragile importance scores

Researchers tested four small reasoning models and found that measuring which parts matter most for pruning can produce unreliable results. Scoring final-answer tokens led to accuracy losses of 27.6 and 32.2 points on two models compared with scoring reasoning tokens, even though perplexity barely changed. The study also found that individual neurons were too small a unit for reliable importance scores, with more consistent results appearing in groups of 100 neurons. The researchers recommend scoring reasoning tokens, measuring importance at a larger scale and judging results by accuracy rather than perplexity.

Exabeam turns Sherpa into an AI guided partner system

Exabeam expanded Sherpa from a virtual Channel Account Manager into an AI guided system for partner learning, sales, marketing and programme management. It brings together certifications, sales coaching, competitive guidance, co-branded campaigns, deal registration and market development funds. Sherpa uses partner progress, activities and registered opportunities to suggest next steps, such as completing training or following up on leads. Planned improvements include deeper pipeline intelligence, support for stalled or at-risk deals, targeted marketing suggestions and more AI-driven sales coaching.

Factory expands Microsoft partnership for autonomous software engineering

Factory expanded its collaboration with Microsoft so companies can run its self-improving software factories inside their Microsoft Azure environments and use existing code, identity and governance systems. Factory Droids can investigate engineering issues, write and test fixes, and open pull requests for human approval. They can also review GitHub pull requests for security issues and run codebase audits. Microsoft Execution Containers limit the files and network destinations a Droid can access, with policies managed through Microsoft Intune. Factory is available in Microsoft Marketplace, and eligible purchases count toward a customer's Microsoft Azure Consumption Commitment.

ARISE helps enterprise AI agents learn from failures

Researchers propose ARISE, a memory system that helps enterprise AI agents improve without changing their model weights, tools, or workflows. It uses the agent or a teacher model to find causes of failures, save fixes as small memories, and combine them into rules. ARISE keeps a rule only when it improves accuracy on held out tasks. Across 14 workflow areas and five models, ARISE with Claude Opus 4.6 improved results over the baseline by 10.9 percentage points on SOP-Bench and 5.8 points on τ²-bench. It also beat GEPA on both tests, while using fewer model calls and output tokens at similar task cost.

Brendan Ahern highlights AI investments and industry growth

On November 19, 2026, KraneShares Chief Investment Officer Brendan Ahern spoke with Kelly Evans on CNBC’s The Exchange about AI companies to watch. He said KraneShares had the chance to invest in Anthropic for its U.S.-listed AI ETF, AGIX, in February 2025. Morgan Stanley estimates that nearly $3 trillion in AI infrastructure investment will flow through the global economy by 2028, with more than 80% of that spending still ahead. Anthropic’s annualized run-rate revenue rose from about $1 billion in early 2025 to $47 billion in May 2026. McKinsey estimates that generative AI could add $17.1 trillion to $25.6 trillion to the global economy each year.

European stocks rise as AI optimism lifts technology shares

European stocks rebounded on Thursday as technology shares gained amid optimism about artificial intelligence. The pan-European Stoxx 600 rose 0.4% to 434.4, Germany’s DAX 30 gained 0.5% to 15,444, and the UK’s FTSE 100 added 0.3% to 7,799. The Stoxx 600 technology index climbed 1.4% to 1,444, making it the strongest sector. The article links confidence in AI to more companies investing in the technology and the growth of new AI applications.

vCluster 5.0 offers self service access to private AI hardware

vCluster Platform v5.0 lets customers set up bare metal GPU servers without asking a provider’s staff to handle every request. Providers retain ownership and control how much capacity each customer receives, while customers select from approved machine types, images, and cluster templates. The platform uses vMetal to manage physical servers and adds tenant boundaries and quotas so customers only see their assigned capacity. Customers can use the hardware for different setups, such as Slurm or Kubernetes clusters. Version 5.0 is not generally available, and providers will need to test whether self service speeds up delivery and reduces support needs.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI infrastructure cloud computing Google Cloud AWS AI chips Micron memory shortage AI investment Tencent workplace AI security data privacy language model pruning AI agents enterprise AI ARISE AI coding tools Anthropic GPU provisioning bare-metal GPUs AI market trends

Comments

Loading...