The Week AI Became a Publicly Traded Commodity
Week of May 18 to 25, 2026 | Issue 7 | For smart professionals catching up on AI
This week
The AI industry stopped pretending it was a research project. OpenAI filed confidentially for an IPO targeting a $1 trillion valuation and a September listing. Anthropic is closing a $30 billion round that would value it at $900 billion, making it more valuable than OpenAI on paper. Nvidia posted $81.6 billion in quarterly revenue and the stock still dropped because Wall Street expected more. Google held I/O and shipped Gemini 3.5 Flash, a personal AI agent called Spark, and a video-generating model called Omni across every product it owns. And an OpenAI model disproved an 80-year-old math conjecture that had stumped every human who tried. The week's message was clear: AI is no longer a bet on the future. It is a financial instrument, priced like one, and traded like one. If you use these tools at work, you are now a line item in someone's IPO prospectus.
🗞️ News you should know
1. OpenAI filed for an IPO targeting $1 trillion and a September listing
OpenAI confidentially filed IPO paperwork with the SEC on or around May 21, according to the Wall Street Journal and Reuters. Goldman Sachs and Morgan Stanley are leading the offering. The company was last valued at $852 billion in its January funding round. Sam Altman reportedly wants to go public by September, which would make it one of the largest public market debuts in history. The filing came days after OpenAI won its lawsuit against Elon Musk, removing the last major legal overhang. What this means for you: if your company pays for ChatGPT Enterprise, your vendor is about to become a public company with quarterly earnings pressure. Public companies optimize for revenue growth every 90 days. Expect price increases, upsell pressure, and new "enterprise tiers" by Q4. If you're on a contract that renews in the fall, lock in your current rate now.
Source: reuters.com | Published: 2026-05-20 Corroboration: wsj.com via engadget.com
2. Anthropic is closing a $30 billion round at a $900 billion valuation
Bloomberg reported on May 22 that Anthropic is set to close its latest funding round as soon as next week, raising over $30 billion at a valuation above $900 billion. That would make Anthropic more valuable than OpenAI on secondary markets. The company was valued at $380 billion just three months ago. Revenue reportedly doubled quarter-over-quarter. The round is oversubscribed. What this means for you: the "which AI vendor should we pick" conversation at your company just got a new data point. Anthropic is now the better-capitalized, faster-growing, and (on paper) more valuable company. If your IT team has been defaulting to OpenAI because "they're the biggest," that argument expired this week. Ask which vendor is growing faster and which has more cash to invest in the features you actually need.
Source: bloomberg via yahoo finance | Published: 2026-05-22 Corroboration: pymnts.com
3. Google I/O shipped Gemini 3.5 Flash, a personal agent, and a video model in one keynote
Google held I/O 2026 on May 19 and announced over 100 features. The three that matter: Gemini 3.5 Flash is now the default model across the Gemini app and AI Mode in Search, beating 3.1 Pro on coding and agentic benchmarks at 4x the speed. Gemini Spark is a personal AI agent that lives inside Gmail, Docs, and Calendar and takes actions on your behalf. Gemini Omni generates and edits video from text, images, and audio in a single model. Google also announced Antigravity 2.0 for agentic development and native Kotlin support in AI Studio. What this means for you: if you use Google Workspace, Gemini Spark is the feature to watch. It is not a chatbot. It is an agent that reads your calendar, drafts replies, and takes actions without you asking. Check your Workspace admin panel this week for the Gemini Intelligence rollout. The first thing worth trying: ask Spark to "summarize what I missed this morning" when you open your laptop. If it pulls from your calendar and email correctly, you have a working daily briefing in 10 seconds.
Source: 9to5google.com | Published: 2026-05-19 Corroboration: cnbc.com
4. An OpenAI model disproved an 80-year-old math conjecture, verified by mathematicians
OpenAI announced on May 20 that an internal general-purpose reasoning model autonomously disproved Paul Erdos's 1946 unit distance conjecture, a problem in discrete geometry that had resisted eight decades of human effort. The proof was verified by external mathematicians, including Fields medalist Tim Gowers, who called it "a milestone in AI mathematics." The model produced an infinite family of counterexamples yielding a polynomial improvement over the conjectured bound. Nature, Scientific American, and New Scientist all covered it independently. What this means for you: this is not a product launch. You cannot buy it. But it is the clearest evidence yet that reasoning models can produce original intellectual work, not just summarize or rephrase existing knowledge. The next time someone in your organization says "AI can only remix what humans already wrote," you have a peer-reviewed counterexample. Worth knowing for the next strategy conversation about where AI fits in your R&D pipeline.
Source: openai.com | Published: 2026-05-20 Corroboration: nature.com
5. Nvidia posted $81.6 billion in quarterly revenue and the stock still fell
Nvidia reported Q1 fiscal 2027 earnings on May 20: $81.6 billion in revenue, up 85% year-over-year, with data center revenue at $75.2 billion. Adjusted EPS was $1.87, beating estimates. The company raised its dividend 25-fold and announced a new buyback. The stock still dropped after hours because Wall Street had priced in even more. Revenue guidance for Q2 was $90 billion, above consensus but not enough to justify the run-up. What this means for you: Nvidia's earnings are the closest thing to a real-time demand meter for AI infrastructure. Revenue up 85% means every major company is still buying GPUs as fast as Nvidia can ship them. If your company is "waiting for AI costs to come down," this quarter says that is not happening in 2026. The companies buying those chips are your competitors, your vendors, and your customers. The infrastructure buildout is accelerating, not plateauing.
Source: investor.nvidia.com | Published: 2026-05-20 Corroboration: cnbc.com
6. Anthropic held Code with Claude London and said most of its own code is now written by Claude
Anthropic's developer conference, Code with Claude, ran in London on May 19. The headline from MIT Technology Review's coverage: "Most software at Anthropic is now written by Claude. Claude has written most of the code in Claude Code." OpenAI, Google, and Microsoft made similar claims the same week. The conference featured live demos of multi-agent coding workflows, MCP server integrations, and a zero-to-deployed app built in 30 minutes using subagents. What this means for you: the companies building AI coding tools are now the companies using AI coding tools to build themselves. This is not marketing. It is a structural shift in how software gets made. If your engineering team is still debating whether to "allow AI in the codebase," the companies building the AI already decided. The question is no longer whether to use it. It is how fast your team adopts it before the productivity gap becomes a hiring disadvantage.
Source: technologyreview.com | Published: 2026-05-21 Corroboration: claude.com/code-with-claude/london