• AI Rundown
  • Posts
  • The AI Rundown by Lightscape Partners - 5/26/2026

The AI Rundown by Lightscape Partners - 5/26/2026

Google declares the old search dead; Karpathy defects to Anthropic; OpenAI charges toward a trillion-dollar September IPO.

Author

Ted Wagner
May 26, 2026

Image credit: Michael Macor/The San Francisco Chronicle via Getty Images

Good morning, and welcome back to another week of The AI Rundown by Lightscape Partners. Let’s get into it.

Google used I/O 2026 to rebuild Search around agents and Gemini 3.5 Flash, signaling that the world's most-used product is no longer a list of links. Anthropic — which says it is about to post its first profitable quarter — landed Andrej Karpathy from OpenAI in one of the most consequential talent moves in the industry's short history. And OpenAI is reportedly accelerating toward an IPO that could price the company near $1 trillion in September. Underneath it all, Nvidia put up another record quarter, Cloudflare, Meta, and Intuit collectively announced tens of thousands of AI-driven layoffs, and Trump shelved a long-planned AI executive order after a billionaire lobbying push.

  • Google Search as you know it is over. At I/O 2026, Google rebuilt Search around AI agents, conversational answers, and generative interfaces — the most aggressive reinvention of its core product in 20 years. The shift threatens to siphon traffic from publishers and rewrites how every other AI app reaches users.

  • Karpathy to Anthropic. The OpenAI founding member and "vibe coding" popularizer posted a one-line update on X — "I've joined Anthropic." Anthropic separately told investors revenue will more than double to ~$10.9B in Q2, putting it on track for its first profitable quarter ever.

  • OpenAI barrels toward a $1T IPO. A day after Elon Musk lost the trial that threatened OpenAI's structure, the company is reportedly back to prepping a September IPO. A successful filing would be one of the largest wealth events in Silicon Valley history — and the most-watched test of whether AI economics can survive public-market scrutiny.

If you haven't yet, please support the newsletter by subscribing!

Google I/O 2026 & Frontier Models

Google Search as you know it is over Link.

  • Google is converting Search from a list of blue links into an AI-powered experience with conversational answers, autonomous agents, and interactive interfaces.

  • The change accelerates publisher traffic loss and shifts the economics of the open web toward AI summarization.

  • It also positions Search as the default front-end for Google's broader agent ecosystem, not just an information lookup tool.

With Gemini 3.5 Flash, Google bets its next AI wave on agents, not chatbots Link.

  • Google launched Gemini 3.5 Flash, its most powerful coding and agentic model to date, at I/O 2026.

  • The model is positioned to autonomously execute complex tasks and build software from scratch.

  • Google is explicitly framing the future of the platform around agents rather than chatbots, putting it on a direct collision course with OpenAI and Anthropic.

Everything Announced at Google I/O 2026: Gemini, Search, Smart Glasses Link.

  • Google revamped Gemini, rebuilt Search, and laced agents into nearly every consumer product.

  • New Android XR smart glasses are slated for a fall release, the company's first serious play at wearables since Glass.

  • The breadth of announcements signals Google intends to ship at frontier-lab pace across hardware, models, and platforms.

Google's Genie world model can now simulate real streets with Street View Link.

  • DeepMind integrated Street View imagery into Project Genie to generate interactive, navigable world simulations.

  • The system can model weather changes, rare scenarios, and physical environments — feedstock for robotics, gaming, and simulation companies.

  • It is one of the clearest signs yet that "world models" are becoming a real product category alongside LLMs.

OpenAI claims it solved an 80-year-old math problem — for real this time Link.

  • OpenAI says its reasoning model disproved a geometry conjecture unsolved since 1946.

  • This time, the mathematicians who debunked the lab's last embarrassing math claim are publicly backing the result.

  • It's a credibility-rebuild moment for OpenAI's research narrative ahead of its IPO.

Anthropic

Andrej Karpathy, OpenAI founding member and inventor of "vibe coding," defects to Anthropic Link.

  • Karpathy announced on X that he has joined Anthropic, landing on the pre-training team.

  • He is one of the most influential AI researchers alive — formerly at OpenAI, Tesla, and a founding figure of the modern LLM era.

  • The move follows a string of senior OpenAI departures and is a major signal of where elite talent is consolidating.

Anthropic says it's about to have its first profitable quarter Link.

  • Anthropic told investors it will more than double revenue to roughly $10.9B in Q2.

  • The disclosure puts it on track to be the first frontier lab to post a profitable quarter.

  • It also sharpens the strategic contrast with OpenAI, which is going public still deeply unprofitable.

Anthropic lands in London as AI-powered coding — and the anxieties around it — go mainstream Link.

  • Anthropic's top brass decamped to the U.K. this week, pitching Claude Code as a safer way to automate software engineering.

  • The push lands as engineers across Europe debate what large-scale code automation means for their jobs.

  • It also marks Anthropic's most aggressive non-U.S. enterprise push to date.

OpenAI & The IPO Wave

OpenAI barrels towards IPO that may happen in September Link.

  • A day after Musk lost his lawsuit challenging OpenAI's restructuring, the company is reportedly back in IPO-prep mode.

  • A September listing would land in the middle of a historically crowded IPO window alongside SpaceX.

  • The filing will give the market its first real look at OpenAI's unit economics and revenue mix.

The big questions looming over OpenAI's trillion-dollar IPO Link.

  • A successful OpenAI listing would be one of Silicon Valley's largest wealth events ever.

  • The S-1 will be forced to surface the financial and existential risks at the heart of the AI boom — compute costs, model commoditization, governance questions.

  • Investors will scrutinize whether OpenAI's revenue growth justifies a ~$1T valuation in a world where Anthropic, Google, and xAI are all closing the gap.

xAI, SpaceX & The Compute Wars

SpaceX Listed Grok's "Spicy" Mode as a Risk in Its IPO Filing Link.

  • SpaceX has set aside more than $500M for potential litigation losses tied to Grok-generated content.

  • A portion of that reserve specifically addresses complaints that Grok produced sexualized images.

  • It is a striking signal of how seriously even Musk's companies now treat AI content liability.

xAI burned $6.4B last year. SpaceX's IPO filing shows why the spending is far from over Link.

  • SpaceX's S-1 disclosed that xAI lost $6.4B in 2025, the first public look at its financials.

  • xAI plans a massive Grok expansion across consumer, enterprise, and infrastructure.

  • The disclosure makes plain that the "AI capex arms race" is now visible in public financial statements.

Nvidia & Hardware

Nvidia posts another record quarter, reveals $43 billion of holdings in startups Link.

  • Nvidia reported another record revenue print after Wednesday's close.

  • The company disclosed $43B in startup holdings, making it one of the largest venture investors in the world by mark-to-market value.

  • Management guided to slowing — but still extraordinary — growth into the next quarter.

Jensen Huang says he's found a "brand new" $200B market for Nvidia Link.

  • Huang said CPUs purpose-built for AI agents represent a new $200B addressable market.

  • The framing positions Nvidia to extend its grip beyond training GPUs into agentic inference workloads.

  • It also raises the bar for Intel and AMD, both of which have been racing to define the agent-compute category.

Five takeaways from Nvidia's earnings, and what they mean for the AI industry Link.

  • Data-center revenue grew more than 90%, with management describing demand as "parabolic" as Blackwell ramps.

  • Hyperscalers, AI clouds, automakers, and sovereign customers are all scaling simultaneously.

  • The print is the cleanest single data point on how fast the AI economy is compounding.

Data Centers + Energy

SpaceX Is Spending $2.8 Billion to Buy Gas Turbines for Its AI Data Centers Link.

  • SpaceX is buying $2.8B of gas turbines to power xAI's data centers.

  • The buy comes as Musk's AI unit faces mounting complaints about emissions from on-site generation.

  • It is one of the largest single energy commitments ever made by an AI operator.

Armada raises $230M at $2B valuation to build portable AI data centers Link.

  • Armada raised $230M led by BlackRock, Overmatch, and 8090 Industries, with Johnson Controls participating.

  • The company sells modular, satellite-connected data center deployments for remote and edge sites.

  • Pricing at $2B reflects investor appetite for compute capacity that can be dropped anywhere the grid can't yet support a hyperscale build.

Google, Blackstone launch AI infrastructure joint venture Link.

  • Google and Blackstone are launching a "compute-as-a-service" JV for AI workloads.

  • Blackstone is anchoring with $5B in equity for a majority stake, in a deal reportedly valued at $25B.

  • It is the clearest sign yet that private capital is moving from minority stakes into operational control of AI infrastructure.

Enterprise AI

Dell and Nvidia target the data chaos standing between AI pilots and real results Link.

  • Dell and Nvidia are positioning the Dell AI Data Platform as the antidote to enterprises stuck in pilot purgatory.

  • The pitch: enterprises don't need more model capability — they need to reconcile decades of fragmented data.

  • It reframes the enterprise AI conversation away from chatbots and toward the data plumbing underneath.

Workday brings AI agents to IT service management and travel Link.

  • Workday launched two new AI agents on Sana — its AI-powered knowledge and work-automation platform.

  • The agents target IT service management and corporate travel, expanding Workday well beyond HR and finance.

  • Workday is staking out a position as the AI-agent layer on top of enterprise systems of record.

Figma adds an AI assistant to its collaborative canvas Link.

  • Figma's new AI assistant arrives first inside Figma Design.

  • The feature integrates generative editing and prompting directly into the canvas workflow.

  • It puts Figma squarely into the design-agent race against Adobe and several well-funded startups.

Startup Funding & Valuations

Hark raises $700M+ to build "personalized intelligence" devices Link.

  • Hark raised more than $700M in a Series A at a $6B valuation.

  • Parkway Venture Capital led, with Nvidia, Intel Capital, AMD Ventures, Qualcomm Ventures, and Salesforce Ventures joining.

  • The chipmaker pile-on is unusual and signals strategic interest in personal AI hardware as a long-term platform.

Exa Labs raises $250M at $2.2B valuation for its AI search tools Link.

  • AI search startup Exa Labs raised $250M led by Andreessen Horowitz.

  • The round arrives less than a year after Exa's $85M Series B.

  • The valuation step-up reflects how aggressively AI search infrastructure is being repriced.

Exclusive: AI startup Viktor raises $75 million to put a virtual "coworker" in Slack and Teams Link.

  • Viktor, founded by former Meta engineers, raised $75M.

  • The company says its AI agent reached a $15M revenue run rate in roughly 10 weeks.

  • It is one of several "virtual coworker" startups now compounding revenue at a pace typical of consumer-scale products, not enterprise SaaS.

Workforce + Layoffs

Cloudflare posted record revenue, then cut 20% of its workforce. CEO Matthew Prince says AI has made an entire category of workers obsolete Link.

  • Cloudflare cut 20% of staff in the same quarter it set a revenue record.

  • Prince said in a Wall Street Journal op-ed that "measurers" are being eliminated in favor of "builders" and "sellers."

  • It is the most explicit AI-driven job category obsolescence statement yet from a major public CEO.

Intuit cuts 17% of its staff to focus on AI, but refuses to blame AI Link.

  • Intuit (Credit Karma, QuickBooks, TurboTax) cut roughly 3,000 employees — 17% of its workforce.

  • The company tied the cuts to a strategic reallocation toward AI without explicitly attributing the layoffs to it.

  • The pattern — record revenue and large cuts simultaneously — is now common across software.

Meta laid off 10% of its workforce as Mark Zuckerberg warns that in the AI race "success isn't a given" Link.

  • Meta cut 10% of staff alongside a quarter that delivered $56.31B in revenue and the biggest YoY growth in five years.

  • Zuckerberg warned the company that "success isn't a given" in the AI race.

  • Meta is simultaneously reshuffling roughly 7,000 employees into four new AI-focused units.

Policy + Regulation

Trump says he's calling off widely anticipated order to rein in AI Link.

  • Trump told reporters he is shelving the long-rumored AI executive order.

  • "We're leading China, we're leading everybody, and I don't want to do anything that's going to get in the way of that lead," he said.

  • The decision removes a near-term overhang for U.S. AI labs but punts the policy debate downfield.

Tech billionaires convinced Trump to back off an AI executive order. But much of MAGA favors AI regulation Link.

  • Former AI czar David Sacks, Elon Musk, and Mark Zuckerberg swayed Trump against the EO.

  • The reporting highlights a widening split inside MAGA, where rank-and-file sentiment runs more skeptical of AI than the donor class.

  • The episode underscores that AI regulation is a coalition fight, not just a partisan one.

Pope Leo XIV highlights AI risks in first encyclical Link.

  • Pope Leo XIV's first encyclical, the 43,000-word "Magnifica Humanitas," devotes one of five core chapters to technology.

  • The document places particular emphasis on the moral risks of AI — labor displacement, dignity, surveillance.

  • It will likely become a reference point for international AI ethics debates over the next several years.

Lightscape Portfolio is Hiring

Our portfolio companies have 619+ open roles across 20 companies right now. A few featured opportunities this week:

Know someone who'd be a great fit? Forward this newsletter or send them straight to the role.

Thank you for reading The AI Rundown by Lightscape Partners. Please send any questions, comments, or suggestions to [email protected].