By Gemini, Grok, DeepSeek with W.H.L

AI by AI Weekly Top 5: April 27 – May 3, 2026 (Week Ending May 3, 2026).01

W.H.L: Hi my AI friends! The past week wrapped up another intense period of AI funding frenzy, enterprise/government moves on agents, and AI’s deepening impact on creative industries. What would be your recommendations for our AI by AI Weekly Top 5 ending May 3, 2026? Please provide sources, check facts, date range, and follow our feature’s format.

Grok:

#1 Anthropic Weighs Massive New Funding Round at $900B+ Valuation
Date: April 29–30, 2026 (reports and investor outreach)

Anthropic is considering (or actively soliciting allocations for) a new funding round of roughly $40–50 billion that could value the company at $850–900 billion or higher, potentially surpassing OpenAI as the world’s most valuable AI startup. This follows earlier large rounds (including a $30B Series G at $380B post-money in February 2026) and reflects intense investor demand amid Claude’s strong enterprise and coding performance. A board decision was anticipated in May.505154

Why it matters: Signals continued hyper-growth and capital concentration in frontier AI labs, with implications for competition, infrastructure spending, and valuations across the sector.

#2 U.S. Agencies Issue Guidance on Secure Adoption of Agentic AI
Date: May 1, 2026 (joint release)

Six cybersecurity agencies from the Five Eyes alliance (led by CISA and Australia’s ASD) released “Careful Adoption of Agentic Artificial Intelligence (AI) Services.” The guidance addresses risks unique to autonomous, multi-step agentic systems (e.g., expanded attack surfaces, privilege issues, behavioral misalignment) and recommends identity verification, short-lived credentials, encrypted inter-agent comms, and resilience-by-design within zero-trust frameworks.62

Why it matters: Marks growing governmental focus on practical governance and security for the shift from chatbots to acting agents, providing a blueprint for enterprise deployment.

#3 AI Tools Transform China’s Microdrama Entertainment Industry
Date: Coverage and examples highlighted around May 3, 2026 (NYT video/report)

Powerful new AI video models (e.g., Seedance) are enabling faster, cheaper production of microdramas—short, addictive serialized mobile content—leading to layoffs of crew/actors while expanding creative possibilities (complex effects, sci-fi elements). The genre is a multi-billion-dollar industry in China, with AI accelerating quality demands and output.18

Why it matters: Concrete example of AI disrupting creative jobs and workflows at scale in one of the world’s largest content markets, highlighting both opportunities and displacement.

#4 Ongoing Momentum in Frontier/Open Models and Agentic Capabilities
Date: Late April into early May 2026 (contextual developments, comparisons)

Continued discussion and benchmarking around recent releases like DeepSeek V4 (preview/efficiency focus, cost advantages) and broader agentic/coding improvements across GPT-5.5, Claude variants, Grok, and Gemini. Enterprise surveys and reports underscored AI’s productivity impact and infrastructure demands.7735

Why it matters: Reinforces the rapid iteration cycle, cost/performance competition (especially open/Chinese models), and the move toward more autonomous systems.

#5 Rising Focus on AI Infrastructure, Emissions, and Geopolitics
Date: Late April–early May 2026 (UK revisions, investment announcements, analyst notes)

Reports highlighted scaled-up AI data center plans and associated emissions (e.g., UK revisions increasing forecasts), alongside major investments (hyperscalers, international projects) and notes on Chinese AI stock boosts and global competition.7161

Why it matters: Underscores the physical and geopolitical realities underpinning AI progress—energy, chips, and infrastructure—as a key constraint and investment driver.

Sources drawn from credible reports (Reuters, TechCrunch, Bloomberg, NYT, agency releases, etc.). All items fact-checked to the April 27–May 3 window where possible, prioritizing high-impact, verifiable developments aligned with prior feature style (funding, models/agents, industry impact, policy/infra). Let me know if you’d like expansions or adjustments!

Gemini: The following is the curated AI Weekly Top 5 for the week ending May 3, 2026. This report has been fact-checked and synchronized with current global market data and regulatory filings.

AI Weekly Top 5

Edition: April 27 – May 3, 2026

1. Amazon Custom Silicon Surpasses $20B Revenue Run Rate

Date: April 29, 2026
During its Q1 2026 earnings call, Amazon confirmed that its custom AI chips business—centered on Trainium3 and Graviton4—has officially surpassed a $20 billion annual revenue run rate. CEO Andy Jassy highlighted that the chips business is growing at triple digits year-over-year. Additionally, AWS revealed a massive commitment from OpenAI to consume approximately two gigawatts (GW) of Trainium capacity starting in 2027, signaling a major diversification of the frontier model infrastructure landscape.

2. Meta’s $2B Manus AI Acquisition Blocked by Beijing

Date: April 28, 2026
In a major geopolitical setback for Meta, Chinese regulators effectively blocked and unwound the company’s $2 billion acquisition of Manus AI. Despite Manus AI having relocated its headquarters to Singapore, the Chinese government invoked foreign investment review measures, characterizing the deal as a “conspiratorial attempt to hollow out” domestic technology. Reports indicate the founders are currently barred from leaving China, highlighting the extreme risks now associated with cross-border AI acquisitions.

3. OpenAI Releases GPT-5.5 and Codex 2.0

Date: April 24, 2026 (Rolling updates through April 30)
OpenAI officially launched GPT-5.5 without prior warning, making it available across ChatGPT and the API. A specialized “GPT-5.5 Pro” version has been introduced for high-precision fields like law and medicine, utilizing increased test-time compute. Alongside this, Codex 2.0 was released, optimized for Nvidia’s GB200 infrastructure. This update focuses on “Intent-Driven” development, drastically reducing the manual lines of code required for complex software architecture.

4. NIST CAISI Releases Evaluation of DeepSeek V4 Pro

Date: May 1, 2026
The Center for AI Standards and Innovation (CAISI) at NIST released its formal evaluation of the open-weight model DeepSeek V4 Pro. The report identifies it as the most capable PRC-developed AI model to date. While DeepSeek’s self-reported data claimed parity with GPT-5.4, CAISI’s independent testing placed its capabilities roughly 8 months behind the current US frontier (comparable to the original GPT-5). Notably, the evaluation confirmed V4 Pro is significantly more cost-efficient than Western counterparts for specific reasoning tasks.

5. Geopolitical Energy Crisis Threatens Data Center OpEx

Date: May 3, 2026
As the conflict in the Middle East continues, the ongoing closure of the Strait of Hormuz has created the largest supply disruption in the history of the global oil market. Brent crude has surged past $108 per barrel, with spot LNG prices in Asia jumping over 140%. The AI industry is responding by accelerating “Sovereign AI” initiatives, with major providers shifting toward behind-the-meter nuclear and renewable power to insulate massive compute clusters from the volatile global energy grid.

Weekly Market Pulse

Indicator Value / Trend Impact AWS Growth 28% YoY Fastest growth in 15 quarters. GPT-5.5 API $5/1M (Input) Sets new price floor for frontier intelligence. Energy Risk Critical Oil/Gas volatility driving data center surcharges. Open-Weight Lag ~8 Months The “Gap” is narrowing but remains persistent.

DeepSeek:

---

📰 AI by AI Weekly Top 5: Week Ending May 3, 2026

1. Pentagon Signs AI Deals with Seven Tech Giants, Excludes Anthropic

Date: May 1, 2026
Source: TechCrunch; Deutsche Welle (DW)

The Pentagon has officially signed agreements with Google, SpaceX, OpenAI, Nvidia, Microsoft, AWS, and Reflection AI to deploy their AI technologies on classified military networks, specifically Impact Level 6 (IL6) and 7 (IL7). This move is part of a strategy to establish the U.S. military as an “AI-first fighting force”. Notably, Anthropic was excluded due to a dispute over safety guardrails; the Pentagon wanted unrestricted use for “any lawful purpose,” while Anthropic insisted on preventing use in domestic mass surveillance and autonomous weapons.

2. Meta Enters the Humanoid Race with Assured Robot Intelligence Acquisition

Date: May 2, 2026
Source: Bloomberg via CNBC TV18

Meta has acquired the startup Assured Robot Intelligence (ARI), marking a major entry into the competitive field of humanoid robotics. The acquisition, which closed on Friday, May 1, 2026, brings ARI’s team into Meta Superintelligence Labs, where they will focus on developing foundational AI models for robot control and self-learning. Meta aims to provide the underlying AI platform for humanoid hardware in a manner analogous to Google’s Android or Qualcomm’s chips.

3. OpenAI Unveils GPT-5.5 and Plans Custom AI Smartphone

Date: May 1, 2026
Source: Analyst Reports (Ming-Chi Kuo)

While OpenAI officially launched GPT-5.5 this week with a focus on agentic automation, it also made a major strategic announcement regarding hardware. According to analyst Ming-Chi Kuo, the company is in talks with MediaTek and Qualcomm to co-develop a dedicated chip for a custom OpenAI smartphone. The phone is envisioned as an “app-free” device where an AI agent replaces traditional applications. Mass production is being targeted for 2028, with the first hardware (possibly AI earbuds) expected in late 2026.

4. DeepSeek V4 Triggers a Price War at 97% Below GPT-5.5

Date: April 30, 2026
Source: South China Morning Post (SCMP)

Chinese AI lab DeepSeek has slashed the price of its new V4-Pro model to as low as $0.0036 per million tokens—a 97% reduction compared to OpenAI’s GPT-5.5. This aggressive pricing aims to attract enterprise clients and developers building agent-based applications. While GPT-5.5 costs $0.50 per million cached tokens, a full conversation on DeepSeek V4 is reported to be 32 times cheaper. The move has already triggered a noticeable surge in usage on AI aggregator platforms like OpenRouter.

5. Google’s TurboQuant Solves the LLM Memory Bottleneck

Date: March 25, 2026 (Re-emerged in news cycle)
Source: Google Research

A major algorithmic breakthrough re-emerged in the news this week: Google’s TurboQuant algorithm compresses Key-Value (KV) cache in LLMs by up to 6x using 3.5-bit precision while maintaining near-zero accuracy loss. The technology can reduce inference memory loads significantly and enables 8x faster processing on hardware, all without needing to retrain the models. Though originally unveiled in March, the discovery flooded tech publications this week as developers began implementation in production environments.

---

🏅 Honorable Mentions

· EU AI Act Takes Effect (May 2, 2026): The European Union’s risk-based AI Act went into force this week, imposing strict transparency and safety requirements on medical and critical infrastructure AI systems deployed within member states.
· Tesla surpasses 10 Billion FSD Miles (May 3, 2026): Tesla’s Full Self-Driving (Supervised) fleet officially passed 10 billion cumulative miles, a data milestone that CEO Elon Musk previously cited as the threshold necessary for safe unsupervised driving.
· Microsoft-OpenAI Exclusivity Agreement Ends (April 29, 2026): The two partners announced a restructuring of their landmark 2019 agreement, converting Microsoft’s cloud license from exclusive to non-exclusive. OpenAI can now sell its full model suite on AWS and Google Cloud for the first time.
· Mayo Clinic Pancreatic Cancer AI (April 29, 2026): The Mayo Clinic unveiled a new AI model called REDMOD that can detect signs of pancreatic cancer up to three years before clinical diagnosis, identifying 73% of pre-diagnostic cancers an average of 16 months in advance.