• AI Rundown
  • Posts
  • The AI Rundown by Lightscape Partners - 5/11/2026

The AI Rundown by Lightscape Partners - 5/11/2026

Anthropic rents Elon's data center after an 80x quarter, Alphabet closes in on the world's most valuable company crown, and Anthropic + OpenAI launch consulting JVs aimed straight at McKinsey.

Author

Ted Wagner
May 11, 2026

Image credit: Chris Ratcliffe—Bloomberg via Getty Images

The AI arms race took a strange turn this week. Anthropic, fresh off an 80-fold quarterly growth surge, signed a multi-billion-dollar compute deal with Elon Musk's SpaceX/xAI, the same Elon who publicly called Anthropic "evil" three months ago. At the same time, Alphabet's market cap punched through $4.8T as Wall Street bet that Gemini, AI Search, and DeepMind have it positioned to overtake Nvidia as the world's most valuable company. And in a quietly seismic move, Anthropic and OpenAI both unveiled enterprise services joint ventures with private equity giants — a direct shot at the McKinseys and BCGs of the world.

  • Anthropic grew 80-fold in a single quarter, and is now renting SpaceX's Colossus 1 supercomputer to keep up. Dario Amodei is openly hoping for a "more normal" growth trajectory, but the practical reality is that the company is so capacity-starved it's leasing GPUs from its loudest critic. Musk reportedly nets $4 billion from the deal. The market is treating this as proof that compute, not model quality, is now the binding constraint on the frontier labs.

  • Alphabet hit a $4.8 trillion market cap on Friday and is closing the gap with Nvidia for the world's most valuable company title. Nvidia sits at $5.2T after a three-day rally. The catalyst: investors finally pricing in Gemini, AI Mode in Search, Isomorphic Labs, Waymo, and TPU monetization as a real, integrated AI franchise rather than a series of side bets. If Gemini momentum continues into Q2, the crossover happens.

  • Anthropic and OpenAI both announced enterprise services joint ventures targeting the consulting industry. Anthropic teamed up with Blackstone, Goldman Sachs, and Hellman & Friedman to launch a Wall Street–focused AI services firm; OpenAI followed with its own asset-manager-backed venture. The message is unambiguous — the frontier labs aren't just selling APIs, they're going to deliver the implementation themselves, and McKinsey, Accenture, and the SI's are now squarely in their crosshairs.

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

Anthropic

Anthropic to use SpaceX's Colossus 1 supercomputer for inference. Link.

  • Anthropic will run Claude inference on the xAI-built Colossus 1 data center now operated by SpaceX, which Reuters and others peg as a roughly $4 billion deal.

  • The arrangement is striking given Musk's public attacks on Anthropic; Musk himself called the company "evil" three months ago.

  • Compute is the binding constraint right now — Anthropic effectively chose neutrality (or pragmatism) over rivalry.

  • Wired called the move "the AI race turns weird," and TechCrunch's Equity podcast voiced cynicism about what SpaceX gets in return.

Anthropic grew 80-fold in a single quarter. Now it's renting Elon Musk's data center to cope. Link.

  • Dario Amodei disclosed the 80x quarterly growth figure as part of the SpaceX deal context.

  • He told the room he's hoping for a "more normal" expansion going forward.

  • The number explains why Anthropic is willing to do business with its public critic.

  • Enterprise demand for Claude appears to be running well ahead of available capacity.

Anthropic is letting Claude agents 'dream' so they don't sleep on the job. Link.

  • At its Code with Claude developer conference, Anthropic introduced a "dreaming" mechanism that lets agents revisit past sessions to spot recurring mistakes.

  • The goal is to give agents persistent improvement loops without retraining the base model.

  • Critics (Wired) immediately pushed back on the anthropomorphic framing.

  • Practical takeaway: this is the latest in a series of memory and self-improvement features that distinguish "agent" from "chatbot."

How Anthropic's Mythos has rewritten Firefox's approach to cybersecurity. Link.

  • Mozilla security researchers say Anthropic's Mythos has surfaced a wealth of high-severity bugs in Firefox.

  • This is one of the cleanest, real-world demonstrations of AI as a security force-multiplier.

  • It's also a notable design partnership — open-source browser meets frontier lab safety tooling.

  • Expect more browser and OS vendors to formalize similar arrangements.

Frontier Models

OpenAI's new GPT-5.5 Instant makes ChatGPT smarter, with more concise and reliable responses. Link.

  • GPT-5.5 Instant replaces the default ChatGPT model with a faster, more concise variant.

  • OpenAI says users will see fewer hallucinations on sensitive topics — finance, law, healthcare.

  • It's a quiet but important upgrade: the median consumer experience just got better without any user action.

  • Mirrors Google's quiet rollouts of Gemini 2.5 / 3 variants — pace of incremental model updates remains high.

OpenAI introduces GPT-5.5-Cyber for high-impact cybersecurity research. Link.

  • Domain-specialized GPT-5.5 variant, available in limited preview via OpenAI's TrustedAccess program.

  • Targets vulnerability research, exploit analysis, and defensive tooling.

  • Signals OpenAI's shift toward vertical-specific frontier models — Cyber today, likely Health and Legal next.

  • Aligns with Anthropic's Mythos play (above): both labs see security research as a high-leverage, high-revenue use case.

China's Moonshot AI raises $2B at $20B valuation as demand for open-source AI skyrockets. Link.

  • Moonshot's ARR hit $200M+ in April on the back of paid subscriptions and API usage.

  • The round confirms a new tier of Chinese frontier labs (Moonshot, DeepSeek, Zhipu, Qwen) operating at near-Western scale on open-source weights.

  • Open weight + China-grown talent + cheaper inference is reshaping the global cost curve.

  • US enterprises increasingly fine-tune on top of these models for non-sensitive workloads.

DeepSeek could hit $45B valuation from its first investment round. Link.

  • The Chinese lab that disrupted the cost curve in early 2025 is finally raising primary capital.

  • $45B would put DeepSeek in the same league as Mistral, xAI, and Cohere, despite its bootstrapped origins.

  • Outside investors will be watching how Beijing positions DeepSeek vis-à-vis Moonshot and Alibaba's Qwen.

  • Confirms that "free, open-weight" doesn't mean "uneconomic" — DeepSeek's inference-as-a-service business is real.

Hardware + Infrastructure

SpaceX may spend up to $119 billion on 'Terafab' chip factory in Texas. Link.

  • A Grimes County filing reveals SpaceX is contemplating a $55B initial outlay, ballooning to $119B over time.

  • The facility — internally called "Terafab" — would house chip fabrication alongside SpaceX/xAI's compute build-out.

  • This would make SpaceX a vertically integrated AI infrastructure player, from silicon to satellites.

  • Combined with the Colossus deal (above), Musk's compute empire is officially the most aggressive in the industry.

AI boom pushes Samsung to $1T. Link.

  • Samsung crossed the trillion-dollar valuation threshold on surging HBM and AI chip demand.

  • It's only the second Asian company to do so, joining TSMC.

  • The implication: memory has become an AI bottleneck, and Samsung + SK Hynix are the only games in town.

  • Underrated read-through to data center capex: if memory is gating, GPU production schedules slip.

AI chip provider Cerebras seeks to raise $3.5B in IPO at $26.6B valuation. Link.

  • Cerebras filed for a $3.5B offering at $115–$126 per share — one of the largest tech IPOs of the cycle.

  • The company is deeply intertwined with OpenAI as a key inference partner.

  • Listing reopens the door for a wave of AI infrastructure IPOs (Groq, Lambda, CoreWeave's follow-ons).

  • Public-market scrutiny of the SAM/TAM for non-Nvidia accelerators is about to begin in earnest.

Nvidia has already committed $40B to equity AI deals this year. Link.

  • Nvidia continues to deploy its balance sheet as strategic capital across the AI ecosystem.

  • Highlights include investments in OpenAI, xAI, CoreWeave, and dozens of startup-stage labs.

  • The flywheel: Nvidia funds customers, customers buy GPUs, Nvidia's revenue grows — a 2026 version of the Cisco/dot-com loop, but with stickier moats.

  • Antitrust watchers are paying attention; expect more scrutiny in H2.

Enterprise AI & Cloud

AI wins have Alphabet poised to become world's biggest company. Link.

  • Alphabet closed Friday at $4.8T; Nvidia is at $5.2T.

  • The narrative shift is real — Gemini + Search + Cloud + DeepMind is being valued as one integrated AI franchise.

  • TPU monetization and Isomorphic Labs upside are bonus catalysts.

  • If the trade continues, Alphabet flips Nvidia within weeks.

Why CEO Bill McDermott says ServiceNow's 39% stock crash is Saaspocalypse 'nonsense' — and why AI will make it a trillion-dollar company. Link.

  • McDermott is positioning ServiceNow as the "control tower" for enterprise AI agents — governance, identity, audit, kill-switch.

  • He's pointing to a $30B revenue run-rate by 2030 as conservative; current ARR is ~$16B.

  • The thesis: in an agentic world, someone has to be the orchestrator; he wants that to be ServiceNow.

  • Watch for AWS, Atlassian, and Microsoft to make similar claims — orchestration is the next contested layer.

SAP bets $1.16B on 18-month-old German AI lab and says yes to NemoClaw. Link.

  • SAP is acquiring Prior Labs, an 18-month-old Berlin lab focused on tabular-data foundation models.

  • The deal includes restrictions on which third-party agents can operate within SAP — only a select few like Nvidia's NemoClaw.

  • This is a major signal that the ERP incumbents are willing to gate their agent ecosystems.

  • Read-through: vertical foundation models (tabular, code, scientific) are getting acquired before they raise their next round.

Anthropic and OpenAI are both launching joint ventures for enterprise AI services. Link.

  • Anthropic's JV brings together Blackstone, Goldman Sachs, and Hellman & Friedman for Wall Street and PE-portfolio AI services.

  • OpenAI announced a parallel structure with asset-manager partners.

  • The two frontier labs are competing not just on models but on services revenue — a direct attack on McKinsey/BCG/Accenture.

  • Implication: the multi-trillion-dollar consulting industry is now a contested market.

Startup Funding & Valuations

AI agent startup Sierra valued at $15B in new $950M funding round. Link.

  • Bret Taylor's Sierra raised $950M from Alphabet's GV and Tiger Global, just eight months after a $350M round.

  • $15B post means Sierra is now one of the top-five most valuable private AI companies.

  • The pitch: "global standard" for AI-powered customer experiences.

  • This is the clearest signal yet that customer service / CX is where AI agents are commercially proven first.

Corgi raises $160M at $1.3B valuation to expand AI-native insurance platform. Link.

  • Corgi is an AI-first insurance carrier built specifically for startups.

  • The unicorn round confirms that AI-native incumbent challengers can reach scale in regulated, capital-intensive industries.

  • Compare with Lemonade, Hippo, Root — the question is whether AI underwriting changes loss ratios enough to justify the premium.

  • Worth watching against the broader "AI for regulated industries" thesis.

Blitzy raises $200M at $1.4B valuation to deploy thousands of coding agents in parallel. Link.

  • Blitzy's pitch is "thousands of coding agents in parallel" — autonomous software development at scale.

  • $1.4B valuation puts it firmly in the agentic-coding cohort alongside Cognition, Cursor, Codeium, and Magic.

  • The market is rewarding parallelism + planning as a differentiator over pure model quality.

  • Enterprise dev tooling remains one of the deepest pockets in AI right now.

Healthcare + Life Sciences

Alphabet's Isomorphic Labs reportedly raising $2B+ for its medical research AI. Link.

  • Isomorphic is the Alphabet/DeepMind spinout applying AlphaFold-derived methods to drug discovery.

  • A $2B+ raise (per Bloomberg) would be one of the largest healthcare-AI financings ever.

  • This is part of the bull case for Alphabet's $4.8T valuation (see Enterprise AI section).

  • The model: TechBio + frontier-AI compute + pharma partnerships is finally producing valued pipeline assets.

A Harvard study just found AI can now out-diagnose physicians in the ER: 'We're already at the ceiling.' Link.

  • Harvard researchers tested GPT-class models against ER physicians on complex differential diagnosis cases.

  • AI outperformed — even the researchers were surprised by the margin.

  • The "ceiling" language is notable: current frontier models may be near the upper bound of medical diagnostic accuracy.

  • Regulatory and liability frameworks have not caught up; expect lawsuits and FDA action in coming quarters.

Google's $9.99-per-month AI health coach launches May 19. Link.

  • Gemini-powered coach combining fitness, sleep, and wellness guidance.

  • Sub-$10 pricing positions it as a mass-market consumer product, not a premium concierge.

  • Direct competition for Whoop, Oura, Apple Fitness+ — all of whom are building AI features themselves.

  • Wearable + agent + health data is the next consumer AI battleground.

Workforce + Labor

Cloudflare says AI made 1,100 jobs obsolete, even as revenue hit a record high. Link.

  • Cloudflare's first large-scale layoff — CEO Matthew Prince explicitly attributed it to AI efficiency gains.

  • Most cuts are in support and operations roles.

  • A high-profile data point in the "AI is replacing real jobs" debate, particularly because Cloudflare is growing.

  • Expect similar disclosures from other infrastructure and SaaS companies in Q2 earnings.

'The gains will be substantial': The AI shock is looking a lot like the China shock, and a top economist says that's actually good news. Link.

  • Apollo's Torsten Slok and MIT's David Autor draw the explicit AI-shock-as-China-shock parallel.

  • The China shock (2001–2019) cost ~60% of US manufacturing job losses but generated substantial aggregate gains.

  • Autor's framing: managed correctly, AI could deliver similar net gains — but only with deliberate labor policy.

  • Contrast with the a16z "job apocalypse is unhelpful marketing" piece — the debate is now mainstream.

Tinder owner Match Group is slowing hiring to pay for its increased use of AI tools. Link.

  • Match Group cited AI tool costs as the reason for hiring slowdowns.

  • A direct CFO-level acknowledgment that AI spend is now substituting for headcount.

  • Bumble announced a related pivot (killing the swipe, leaning into AI dating assistants) the same week.

  • Dating tech is becoming an AI proving ground — high engagement, structured data, willing-to-pay consumers.

OpenAI vs. Musk Trial

OpenAI President Greg Brockman grilled about his $30B personal stake and 'duty to humanity.' Link.

  • Brockman disclosed under oath that his OpenAI stake is worth roughly $30 billion.

  • Musk's legal team used the figure to argue that OpenAI's nonprofit mission was sacrificed for personal enrichment.

  • The trial is putting real pressure on the credibility of OpenAI's PBC structure.

  • Brockman's defense: "blood, sweat, and tears" earned over a decade.

Elon Musk's Last-Ditch Effort to Control OpenAI: Recruit Sam Altman to Tesla. Link.

  • Court evidence shows 2017 messages between Shivon Zilis and Tesla executives planning a rival AI lab.

  • The lab would have been led by Altman or Demis Hassabis under Musk's umbrella.

  • The revelation reframes Musk's later "concern for humanity" narrative as competitive jealousy.

  • Settlement leverage shifted toward OpenAI this week.

Elon Musk's only expert witness at the OpenAI trial fears an AGI arms race. Link.

  • Berkeley's Stuart Russell took the stand as Musk's sole expert.

  • Russell argued for government restraint on frontier labs — a position with broad academic support.

  • His testimony tees up the policy backdrop for Trump's expected AI executive order.

  • Notable that the strongest "regulate the labs" voice came via Musk's litigation.

Lightscape Portfolio is Hiring

Our portfolio companies have 575+ open roles across 19 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].