This Week in AI: May 22–29, 2026
IPOs, a papal encyclical, a new Claude model, AI reshaping the global workforce and the numbers proving the infrastructure boom is very real.
On this page+
- 1. OpenAI Files Confidential IPO Targeting $1 Trillion Valuation
- 2. Anthropic Becomes the World's Most Valuable Private AI Startup
- 3. Anthropic Launches Claude Opus 4.8
- 4. Pope Leo XIV Releases First Papal Encyclical on AI
- 5. NVIDIA Drops LocateAnything and Jensen Huang Heads to Taipei
- 6. Enterprise AI Hits Mass Scale
- 7. Dell Posts $43.8B Quarter AI Server Orders Hit $24.4B
- 8. AI Reshapes the Global Workforce
- 9. Google's AI Search Sparks a DuckDuckGo Revolt
- 10. Microsoft & Apple: Platform Moves
1. OpenAI Files Confidential IPO Targeting $1 Trillion Valuation
On May 22, OpenAI filed a confidential S-1 registration statement with the U.S. Securities and Exchange Commission, officially kicking off its path to becoming a public company. Goldman Sachs and Morgan Stanley are leading the deal, with September 2026 as the early target listing date. Private investors currently value the company at $852 billion; at listing, analysts expect it could cross $1 trillion the largest tech IPO in history.
The filing remains sealed until approximately 15 days before the investor roadshow. OpenAI does not expect to be profitable until around 2030, with internal projections showing a $14 billion operating loss for 2026.
Source: OpenAI IPO — Confidential S-1 Filed, $1 Trillion Valuation Target
2. Anthropic Becomes the World's Most Valuable Private AI Startup
Anthropic closed a $30 billion+ funding round at a pre-money valuation above $900 billion surpassing OpenAI's $852 billion March 2026 valuation for the first time. The round is co-led by Sequoia Capital, Dragoneer, Altimeter Capital, and Greenoaks Capital Partners, each investing approximately $2 billion.
The company simultaneously revealed explosive financial momentum: Q2 2026 revenue is projected at $10.9 billion a 130% jump from Q1's $4.8 billion with a projected $559 million operating profit, which would mark Anthropic's first-ever profitable quarter.
Adding fuel to the story, SpaceX's IPO filing disclosed that Anthropic is paying $1.25 billion per month for access to over 220,000 NVIDIA GPUs across SpaceX's Colossus and Colossus II clusters. Elon Musk has since clarified the deal is a 180-day rolling lease rather than a fixed multi-year commitment.
Sources:
3. Anthropic Launches Claude Opus 4.8
On May 28, Anthropic released Claude Opus 4.8 available immediately on claude.ai, the Claude API, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry. The new model arrives just 41 days after Opus 4.7.
Key improvements include sharper judgment, greater honesty about its own limitations, and the ability to work independently for longer on complex tasks. Opus 4.8 is four times less likely than its predecessor to leave flaws in generated code unremarked.
Alongside the model, Anthropic launched dynamic workflows (running multiple subagents in parallel) and a new effort control panel letting users configure how much compute Claude applies to a response. Anthropic teased that Mythos-class models will follow "in the coming weeks."
Sources:
4. Pope Leo XIV Releases First Papal Encyclical on AI
On May 25, exactly 135 years after Pope Leo XIII signed Rerum Novarum on workers' rights during the Industrial Revolution Pope Leo XIV released Magnifica Humanitas ("Magnificent Humanity"), the first papal encyclical ever written about artificial intelligence.
The ~42,000-word document calls for the protection of human dignity in the age of AI, condemns autonomous weapons systems, and argues that "technology is never neutral." Leo presented the encyclical personally a rare departure from tradition with Anthropic co-founder Chris Olah in attendance. The Vatican's decision to partner with Anthropic rather than Google or OpenAI was widely read as a deliberate statement on AI safety values.
Sources:
- Vatican Primary Source — Presentation of Magnifica Humanitas
- Pope Leo XIV's Encyclical — Vatican News
5. NVIDIA Drops LocateAnything and Jensen Huang Heads to Taipei
On May 26, NVIDIA released LocateAnything a vision-language model for fast, high-quality visual grounding that enables precise object localisation across images, GUIs, documents, robotics environments, and autonomous driving scenes.
Its core innovation is Parallel Box Decoding (PBD), which predicts complete bounding box coordinates in a single parallel step rather than token-by-token, delivering up to 2.5× higher throughput over prior methods. The model was trained on 12 million images, 138 million+ queries, and 785 million bounding boxes. The 3B-parameter variant is available now on Hugging Face.
Meanwhile, CEO Jensen Huang touched down in Taipei on May 23 ahead of Computex (June 1–5), confirming that Vera Rubin NVIDIA's next-gen AI platform delivering 3.5× training and 5× inference performance over Blackwell is in full production. Huang also teased "a surprise new product we haven't told anyone about yet" arriving in H2 2026.
Sources:
- LocateAnything — NVIDIA Research
- LocateAnything-3B — Hugging Face
- Jensen Huang at Computex — TechTimes
6. Enterprise AI Hits Mass Scale
KPMG deployed Claude to its entire 276,000-person global workforce across 138 countries via its Digital Gateway platform, with an initial focus on tax and private equity clients.
OpenAI launched DeployCo a $4 billion consulting subsidiary at a $10 billion pre-money valuation, led by TPG with McKinsey, Bain & Company, and Capgemini as partners. DeployCo acquired applied AI consultancy Tomoro, seeding 150 forward-deployed engineers from day one.
Sources:
- KPMG & Anthropic Global Alliance — Anthropic
- KPMG Official Press Release
- OpenAI DeployCo Launch — OpenAI
7. Dell Posts $43.8B Quarter AI Server Orders Hit $24.4B
Dell Technologies reported record quarterly revenue of $43.8 billion on May 28, up 88% year-over-year, driven almost entirely by AI infrastructure demand. The company booked $24.4 billion in AI orders and recognised $16.1 billion in AI server revenue in a single quarter. Dell raised its full-year AI server revenue guidance to $60 billion for FY27.
The results are the clearest single-company signal yet that enterprise spending on AI infrastructure is not slowing down.
Sources:
8. AI Reshapes the Global Workforce
Meta moved 7,000 employees into four new AI-focused units: Applied AI Engineering, Agent Transformation Accelerator, Central Analytics, and Enterprise Solutions — while simultaneously laying off 8,000 workers, affecting roughly 20% of its total workforce.
Standard Chartered announced plans to cut 7,000+ back-office roles by 2030, replacing what CEO Bill Winters called "lower-value human capital" with AI and automation. Regulators in Hong Kong and Singapore have since sought clarification from the bank following Winters' remarks. The bank is targeting a 15% return on tangible equity by 2028 and 18% by 2030.
Sources:
- Meta Shifts 7,000 Employees into AI — NBC News
- Meta Layoffs and AI Reorganisation — SiliconAngle
- Standard Chartered 7,000 Job Cuts — Tom's Hardware
9. Google's AI Search Sparks a DuckDuckGo Revolt
Google's replacement of traditional search results with AI agents triggered a measurable backlash DuckDuckGo reported a 30% spike in app installs from users explicitly rejecting being "force-fed" AI-generated answers. The privacy-first search engine is positioning itself as the non-AI default for users who want to find things, not be told things.
Source: DuckDuckGo Installs Up 30% — TechCrunch
10. Microsoft & Apple: Platform Moves
Microsoft made computer-using agents in Copilot Studio generally available, allowing enterprises to build agents that interact with any website or desktop application through the UI no API required. Ships with Claude Sonnet 4.5 and OpenAI CUA as supported models, Azure Key Vault credential storage, and Microsoft Purview audit logging.
Apple quietly registered genai.apple.com ahead of WWDC 2026 (June 8), signalling a major generative AI push in iOS 27. Bloomberg reports iOS 27's rebuilt Siri will run on a custom model based on Google's Gemini technology, processed through Apple's Private Cloud Compute. The domain goes live on June 8.
Sources:
