May was the month the AI race stopped being about who has the smartest model and started being about who can deliver capable models fast, cheap, and reliably enough to put in production. The frontier labs spent the month competing on agents, coding, and price — while the talent war and the regulatory backdrop both shifted underneath them. Here’s what actually moved, and what it means if you’re deciding where AI fits your roadmap.
Google Goes Agentic with Gemini 3.5
At I/O on May 19–20, Google launched Gemini 3.5, leading with Gemini 3.5 Flash — now the default across the Gemini app, Search’s AI Mode, and the API. The pitch was explicit: frontier-class coding and agentic performance at Flash-tier speed and cost, with Google claiming roughly 4x the output throughput of competing frontier models.
Why it matters for builders: Google is no longer playing catch-up on raw capability; it’s competing on the economics of running agents at scale. If your product makes high-volume model calls — support automation, document processing, code generation — a fast, cheap default model changes your unit economics, not just your demo. Re-run your cost models before assuming last quarter’s provider is still the right one.
OpenAI Ships GPT-5.5 Instant — and Targets Trust
On May 5, OpenAI made GPT-5.5 Instant the default ChatGPT model. The headline wasn’t a benchmark leap; it was reduced hallucination in high-stakes domains like law, medicine, and finance, while keeping low latency (and, notably, cutting the gratuitous emoji).
Why it matters for operators: The interesting signal is the targeting. The labs have figured out that the blocker to enterprise adoption isn’t intelligence — it’s reliability in domains where a confident wrong answer is expensive. If you’ve been holding AI back from regulated or high-liability workflows, the reliability gap is narrowing. The judgment call is still yours: narrow the scope, keep a human in the loop, and measure error rates on your own data before you trust the vendor’s.
Anthropic Hits a $965B Valuation and Files to Go Public
On May 28, Anthropic closed a $65B Series H at a $965B post-money valuation, eclipsing OpenAI and confidentially filing for an IPO. The company cited a revenue run rate that had climbed past $45B.
Why it matters: This is the clearest sign yet that the enterprise AI budget is real and durable, not a hype cycle. For builders, the practical read is stability — the model you build on this year is backed by a company that isn’t going anywhere. But concentration is the flip side. Architect for portability: keep an abstraction layer between your application and any single model provider so a pricing or policy change doesn’t become a rewrite.
The Talent War Gets a Marquee Hire
On May 19, Andrej Karpathy joined Anthropic’s pre-training team, leaving his AI-education venture to get back to frontier R&D. It’s a symbolic coup in an increasingly brutal competition for elite researchers.
Why it matters: Where the best people go is a leading indicator of where capability accelerates next. You don’t need to chase headlines, but the signal informs vendor bets: the labs attracting top pre-training talent are the ones most likely to widen their lead over the next 12 months.
The EU AI Act Blinks
On May 7, EU lawmakers reached political agreement on a “Digital Omnibus” that extends compliance deadlines for high-risk AI systems and simplifies parts of the AI Act, while adding targeted new rules.
Why it matters for operators: If EU compliance was a reason to stall, you just got breathing room — but not a reprieve. Transparency obligations for chatbots still arrive in August, and fines remain steep. Use the extra runway to build governance in deliberately rather than scrambling later.
The throughline this month: capable AI is getting cheaper and more reliable faster than most roadmaps assume. The advantage now goes to teams that can tell which shifts actually change their build-vs-buy math — and act on them without over-rotating. If you want a clear-eyed read on which of these moves should change your plan this quarter, let’s talk.