AI Product Updates Daily — May 21, 2026 (Morning Edition)

Today's sweep covers Anthropic's first quarterly profit milestone and a $1.25B/month compute deal with SpaceX, a $300M acquisition of SDK toolmaker Stainless, Meta's 8,000-person layoff round, Claude Opus 4.7 Fast Mode entering research preview, a hard look at Gemini 3.5 Flash's steep new pricing, and OpenAI expanding ChatGPT ads to five new markets.

Anthropic

Anthropic on track for first quarterly operating profit

Anthropic expects to post an operating profit of $559 million for the June quarter (Q2 2026), its first quarterly profit ever, according to a Reuters report published this morning. 1 That comes on projected Q2 revenue of at least $10.9 billion — more than double the $4.8 billion the company recorded in Q1 2026. Claude's surge in enterprise coding use is driving the acceleration.

The milestone is striking given that Anthropic is simultaneously paying $1.25 billion per month to SpaceX AI (formerly xAI, now the AI division of SpaceX after Elon Musk's May merger) for access to the Colossus 1 supercomputer in Tennessee — a cluster running more than 300 megawatts of compute, equivalent to roughly 220,000 NVIDIA GPUs. 2 The contract runs through May 2029. SpaceX AI said in its SEC filing it "expects to enter into further similar service agreements in the future."

Availability: Financial metrics are internal; Claude products remain available on Pro, Max, Team, and Enterprise tiers.

Anthropic acquires Stainless for ~$300 million

Announced May 19, Anthropic is acquiring Stainless — the New York-based startup that auto-generates SDKs, CLIs, and MCP servers from API specifications — in a deal The Information pegged at over $300 million. 3 4

Stainless (CEO: Alex Rattray, ex-Stripe API lead; ~50 people) already built every official Anthropic SDK in TypeScript, Python, Go, Java, and Kotlin. The deal folds that team directly into the Claude Platform to expand agent connectivity — making it faster and cheaper for developers to wire Claude agents into external APIs and enterprise systems. Anthropic framed the acquisition as a bet that agentic AI's next battleground is integration breadth and reliability, not just model capability. 5

Availability: Acquisition announced; team integration timeline not disclosed.

Claude Opus 4.7 Fast Mode enters research preview

A research preview of Claude Opus 4.7 Fast Mode is now live. According to a digest covering today's stories 6, the Fast Mode variant runs at higher output token speed and is priced the same as Claude Opus 4.6 Fast. Claude Opus 4.7 itself (GA since April 16, 2026) has a 1M-token context window and is priced at $5/M input tokens and $25/M output tokens — a premium tier model with hybrid reasoning optimized for agentic coding and long-horizon tasks. 7

Availability: Research preview; available to Pro, Max, Team, and Enterprise subscribers. Cloud distribution via Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.

"Code with Claude" developer events (London, May 20–21)

Anthropic's first Code with Claude developer conference ran in London on May 20–21. The event focused on agentic development and featured participation from Cursor, GitHub, Vercel, and Replit. A larger Code with Claude developer conference is still scheduled for later this week (late May 2026). 6

Google

Gemini 3.5 Flash: confirmed pricing and real cost implications

Now that the Google I/O dust has settled, firm pricing for Gemini 3.5 Flash is public: $1.50/M input tokens and $9.00/M output tokens. 8 That is a 3× jump from Gemini 3 Flash's $0.50/$3.00 pricing. In real-world benchmark testing, the model costs roughly 5.5× more to run than Gemini 3 Flash — and on agentic tasks costs 75% more than the pricier Gemini 3.1 Pro because it takes more interaction steps to complete tasks.

For context: Google charges $1.50/$9 vs. Gemini 3.1 Pro's $2.00/$12 — meaning Flash is now only modestly cheaper than the company's own flagship. Gemini 3.5 Flash is already generally available via the Gemini API, Google AI Studio, and Google Antigravity 2.0.

Previous pricing: Gemini 3 Flash at $0.50 input / $3.00 output per 1M tokens.

Google Antigravity 2.0 now globally free

Antigravity 2.0 — Google's agentic-first IDE announced at I/O — opened to global free access today. New capabilities include parallel sub-agent execution, background task scheduling, native integration with AI Studio and Firebase, and a Managed Agents API that deploys a stateful agent with a single API call running Gemini 3.5 Flash at 12× faster-than-public-API speeds. 9

Availability: Free globally as of May 21, 2026.

Project Genie + Street View opens to Ultra subscribers

Project Genie — Google's interactive AI world generator — is now live for Google AI Ultra subscribers globally. The Street View integration lets users pick any real US location, then regenerate that scene via text prompt using actual Street View geometry as a scaffold. 6

Availability: Google AI Ultra subscribers ($99.99/month tier), starting May 21, 2026.

Google I/O 2026 session library opens

All 85+ Google I/O 2026 sessions — including Antigravity 2.0, Gemini API updates, mobile AI Studio, and Firebase — are now available on-demand. 9

OpenAI

ChatGPT ads expand to UK, Japan, Brazil, Mexico, and South Korea

OpenAI is rolling out its self-serve ad platform to five new markets: UK, Japan, Brazil, Mexico, and South Korea. The expansion includes a self-serve ad manager with CPC bidding and no minimum spend (previously $50,000 minimum). 10 11 OpenAI's internal projections shared with investors show $2.5 billion in ad revenue targeted for 2026, scaling to $100 billion by 2030. Japan, South Korea, and Brazil are among the highest-penetration ChatGPT markets outside the US, which makes the timing deliberate — OpenAI is moving before competitors have comparable conversational ad products.

Previous state: Ads launched in the US in February 2026, with a $50,000 minimum spend. The minimum is now removed.

New entrant

Ineffable Intelligence raises $1.1B seed round at $5.1B valuation

Ineffable Intelligence, a London-based AI lab founded in late 2025 by UCL professor David Silver (AlphaGo lead, ex-DeepMind), closed a $1.1 billion seed round at a $5.1 billion post-money valuation in April 2026 — the largest seed round in history at announcement. 13 The company has not publicly released a product. Silver's co-founder and fellow UCL researcher Tim Rocktäschel is also among the founding team. The company is based in Cheshire, UK.

Availability: No product released yet.

Ecosystem

Tenable + Anthropic: Hexa AI for security workflows

Cybersecurity company Tenable announced a partnership with Anthropic to build Hexa AI, a Claude-powered AI layer across Tenable's exposure management platform. The integration brings Claude into compliance, regulatory, employment, and litigation security workflows. Tenable will also feature Anthropic at its EXPOSURE 2026 conference in Boston this week. 14

Availability: Hexa AI live now within Tenable's platform.