DeepSeek Is Raising a Giant Round and Planning to Go Public, Perhaps This Year

---

1. Key Themes

Open-Source AI Is Winning the Volume War Against Frontier Models

Chinese open-weight models are capturing developer mindshare at a scale that challenges the dominance of closed, frontier labs like OpenAI and Anthropic.

"Chinese open-weight models accounted for 41% of downloads on Hugging Face this spring, surpassing U.S. models. On OpenRouter, the top six most popular models are all open models from Chinese firms, including Tencent, Xiaomi, DeepSeek, MiniMax, and Z.ai. Anthropic's Claude Opus 4.7 trails in seventh place."

"Open models handled nearly a third of AI requests on the [Vercel] platform in June."

---

The Strategic Case for Model Ownership Over API Dependency

Enterprise customers are reconsidering their reliance on closed AI APIs, driven by cost, control, and strategic risk concerns.

"If you're an AI company or a technology company, you don't want to outsource your core capabilities to another company, to a black box API that you don't control, don't have any visibility on, and don't really have any sort of ownership." — Hugging Face CEO Clem Delangue

"Hugging Face's customers and community members are increasingly touting the benefits of owning their own AI models rather than renting them, a trend that's picked up steam in the cold light of day after getting the bill associated with the cost of scaling closed frontier models."

---

DeepSeek's Potential IPO as a Geopolitical and Competitive Inflection Point

A DeepSeek IPO at a $71B valuation would validate the Chinese open-source AI model as a legitimate long-term competitive threat — not just a one-time shock.

"Chinese LLM startup DeepSeek is preparing for an IPO – potentially as soon as later this year – while seeking to raise about $1.5 billion at a $71 billion valuation. DeepSeek has grown fast by offering lower-cost, open-source AI models, including ones that run on Huawei chips rather than U.S.-made chips that are subject to export controls."

---

AI Infrastructure Is Facing a Regulatory and Environmental Reckoning

Both regulatory action and corporate compliance failures signal that AI compute buildout is running ahead of legal and environmental guardrails.

"New York Gov. Kathy Hochul signed the nation's first statewide moratorium on hyperscale data centers, pausing approvals for projects using 50 megawatts or more for one year while the state studies their impact on energy demand, water use, air quality, and utility bills."

"Reuters is reporting that xAI installed 59 unpermitted natural gas turbines for its Colossus 2 data center project near Memphis – more than double the 27 turbines the company previously acknowledged – with potential pollution far above federal permitting thresholds."

---

AI Token Spend Is Becoming a Material Operating Cost for Tech Companies

AI usage is no longer a rounding error — it is approaching the scale of headcount costs and must be managed as a core operating line item.

"Instagram head Adam Mosseri says Meta may eventually cap AI token spending by engineer, warning that heavy AI usage could soon cost as much as an employee's salary and will need to be managed like payroll, GPU capacity, or other operating expenses."

---

2. Contrarian Perspectives

Frontier Models May Become Niche Tools, Not the Default Backbone of AI

The consensus view is that frontier models (GPT-5, Claude 4, etc.) are the center of gravity for AI. The data increasingly suggests the opposite — they will serve only the highest-value, most specialized workloads.

"Maybe in a few years, the frontier models will be for experimenting and [for] some really high-value tasks, and most of the production workloads will actually be powered either by private models within companies or by open source models." — Hugging Face CEO Clem Delangue

The implication for investors: the massive valuations of closed-model labs may be premised on becoming the default infrastructure layer, a position they are losing to Chinese open-weight alternatives.

---

OpenAI's Ad Business Is a ~$100B Fiction

OpenAI projected $100 billion in ad revenue. Independent estimates suggest this is off by an order of magnitude — creating a significant valuation risk for the company and its investors.

"OpenAI's ad business is on pace to miss the company's own five-year forecast by 90%, with Emarketer estimating the entire chatbot ad market will reach just $5.41 billion by 2030, far below OpenAI's projected $100 billion in ad revenue."

---

Only 28% of Venture-Backed Companies Clear the Rule of 40 — and Profitability Is Almost Never Why

The widespread assumption that mature SaaS companies can reach the Rule of 40 via either growth or profitability is contradicted by data: the vast majority who qualify do so exclusively through growth.

"Only 28% [of 1,377 venture-backed companies with $1M+ in revenue] cleared the Rule of 40, and 89% of those did it through growth, not profitability." (via Standard Metrics/sponsored section)

---

3. Companies Identified

DeepSeek | Chinese LLM startup | Preparing IPO at $71B valuation, raising $1.5B; offers lower-cost open-source AI models running on Huawei chips; top model on OpenRouter

"DeepSeek has grown fast by offering lower-cost, open-source AI models, including ones that run on Huawei chips rather than U.S.-made chips that are subject to export controls."

Hugging Face | AI model hosting and developer community platform | Case study for the rise of open-source model adoption among enterprises

"Hugging Face's customers and community members are increasingly touting the benefits of owning their own AI models rather than renting them."

OpenAI | Frontier AI lab | Featured for leaked hardware device plans, ad revenue shortfall, and as investor in Chai Discovery

"OpenAI's first hardware device is reportedly a screenless, mobile smart speaker designed as a 'humanlike AI companion' for the home, with moving parts, ChatGPT integration, and the ability to personalize responses using information from a user's digital life."

xAI | Elon Musk's AI company | Cited for regulatory compliance failure at its Memphis data center

"xAI installed 59 unpermitted natural gas turbines for its Colossus 2 data center project near Memphis – more than double the 27 turbines the company previously acknowledged – with potential pollution far above federal permitting thresholds."

Chai Discovery | AI drug molecule design startup, 2 years old, San Francisco | Raised $400M at $3.8B valuation; notable as a convergence of AI + biotech with top-tier investor syndicate including OpenAI

"Chai Discovery...develops AI models that help drugmakers design molecules by predicting and reprogramming how molecules interact."

Neko Health | Preventive health scan startup, Stockholm | Raised $700M Series C; co-founded by Spotify's Daniel Ek; attracting celebrity investors

"Neko Health...offers full-body preventive health scans using proprietary sensors, blood analysis, and clinician consultations."

Switch | Data center infrastructure, Las Vegas | Planning IPO targeting $80B valuation; beneficiary of AI compute demand

"Switch...has hired Goldman Sachs and JPMorgan for an IPO that could raise up to $10 billion and value the company at nearly $80 billion."

Uber | Ride-hailing / delivery / AV platform | Featured for its coopetition dynamics with Waymo and potential Delivery Hero acquisition

"The company that started as a ride-hailing app now juggles autonomous vehicles, delivery, hotel bookings, and a fleet of robotaxi partners it also happens to compete with."

Brinc | Emergency response drone startup, Seattle | Raised $125M led by Motorola Solutions; a signal of drone adoption in public safety

"Brinc...builds drones for police, fire, and other emergency-response agencies."

LimX Dynamics | Humanoid robot startup, Shenzhen | Raised $200M at $2.2B valuation; signals continued Chinese investment in physical AI

"LimX Dynamics...develops autonomous humanoid robots for commercial service work and entertainment deployments."

Pearl Health | AI-powered value-based Medicare care company, New York | Raised $110M including $50M equity from a16z; notable convergence of AI + healthcare reimbursement reform

"Pearl Health...uses AI to help healthcare providers manage risk and deliver value-based care for Medicare patients."

Singularity | Counter-weapons startup, El Segundo | Raised $80M Series A; defense tech with automotive-scale manufacturing thesis

"Singularity...builds counter-weapons for military users with automotive-style manufacturing designed to increase production volumes."

State Affairs | Legislative tracking platform, Washington DC | Raised $70M from Founders Fund and Khosla; govtech meets AI

"State Affairs...helps lawmakers, agencies, and companies track legislation, regulations, hearings, and policy trends using statehouse reporting and government data."

TerraFirma | Autonomous construction equipment startup, Austin | Raised $115M from Kleiner Perkins and Bain Capital Ventures; physical-world automation

"TerraFirma...develops remote-operated and semi-autonomous construction equipment."

Greylock | 61-year-old Menlo Park VC firm | Raised $1.5B for its 18th fund; continued confidence in early-stage AI bets

"Greylock...backs early-stage startups across AI infrastructure, cybersecurity, applications, deep tech, consumer, and fintech."

Delivery Hero | Global food delivery company | In advanced acquisition talks with Uber; potential antitrust implications

"Delivery Hero confirmed it is in advanced talks with Uber over a potential takeover offer, a deal that would expand Uber Eats across Europe, the Middle East, Asia, and Latin America."

---

4. People Identified

Clem Delangue | CEO, Hugging Face | Central voice in the open-source vs. frontier model debate; articulating the enterprise shift toward model ownership

"Maybe in a few years, the frontier models will be for experimenting and [for] some really high-value tasks, and most of the production workloads will actually be powered either by private models within companies or by open source models."

Adam Mosseri | Head of Instagram, Meta | Flagging AI token spend as an emerging cost management challenge comparable to headcount

"Meta may eventually cap AI token spending by engineer, warning that heavy AI usage could soon cost as much as an employee's salary and will need to be managed like payroll, GPU capacity, or other operating expenses."

Demis Hassabis | CEO, Google DeepMind | Calling for an independent, FINRA-like standards body to review frontier AI models pre-release

"Hassabis is calling for an independent standards body to review frontier AI models before release, arguing that an industry-funded, government-backed regulator modeled on FINRA would be more technical, transparent, and consistent than the Trump administration's ad hoc model reviews."

Miles Wang | Researcher, OpenAI (departing) | Leaving OpenAI to found an AI drug discovery startup; may launch with $200M in funding, signaling high-value talent migration from labs to biotech AI

"Miles Wang...is leaving the ChatGPT maker to launch a new startup focused on developing AI models for drug discovery...he may hit the ground running with a $200 million round of funding."

Daniel Ek | Co-founder, Spotify; Co-founder, Neko Health | Raised $700M Series C for preventive health scan company; signals founder-led consumer health as a viable venture category

"Neko Health, an eight-year-old Stockholm startup co-founded by Spotify co-founder Daniel Ek that offers full-body preventive health scans."

Kathy Hochul | Governor, New York | Signed first statewide hyperscale data center moratorium; key regulatory risk factor for AI infrastructure investors

"New York Gov. Kathy Hochul signed the nation's first statewide moratorium on hyperscale data centers, pausing approvals for projects using 50 megawatts or more for one year."

Sachin Kansal | Chief Product Officer, Uber | Articulating Uber's coopetition strategy with autonomous vehicle partners

"We asked Chief Product Officer Sachin Kansal to untangle it all — how Uber manages its complicated, coopetition-fueled relationship with Waymo, keeps its 10 million human drivers happy."

Dylan Field | Co-founder, Figma | Participating investor in Brinc's $125M round; notable angel bet on drone/defense tech

(Mentioned as an individual investor alongside Index Ventures in Brinc's round)

---

5. Operating Insights

Model Ownership as a Strategic Moat — Build vs. Buy Is Now the Core AI Decision

For operators building AI-native companies, dependence on closed API providers is increasingly seen as a strategic liability. Enterprises are beginning to treat proprietary or open-source model ownership the same way they would treat owning their own data infrastructure.

"If you're an AI company or a technology company, you don't want to outsource your core capabilities to another company, to a black box API that you don't control, don't have any visibility on, and don't really have any sort of ownership." — Clem Delangue, Hugging Face CEO

Tactical implication: For founders and operators, budget modeling should account for a "model stack" decision — the marginal cost of closed APIs at scale vs. the capex of fine-tuning or hosting open-weight alternatives.

---

Token Spend Is the New Headcount — Manage It Accordingly

AI usage costs are crossing the threshold from software expense to labor-equivalent cost center. Companies that treat token spend like discretionary SaaS spend will face budget surprises.

"[Adam Mosseri warns] that heavy AI usage could soon cost as much as an employee's salary and will need to be managed like payroll, GPU capacity, or other operating expenses."

Tactical implication: Operators should instrument per-engineer and per-product-line token consumption now, before it becomes unmanageable. Budget for AI spend in headcount planning cycles, not software procurement cycles.

---

6. Overlooked Insights

The Regulatory Risk to Data Center Development Is Now a Nationwide Signal, Not a Local Story

New York's moratorium on hyperscale data center approvals is the first statewide action of its kind in the U.S. This is a leading indicator — not an isolated event — that AI infrastructure buildout will face increasing state-level friction on energy, water, and air quality grounds.

"New York Gov. Kathy Hochul signed the nation's first statewide moratorium on hyperscale data centers, pausing approvals for projects using 50 megawatts or more for one year while the state studies their impact on energy demand, water use, air quality, and utility bills."

For investors in data center REITs, AI infrastructure funds, or hyperscaler equities, this introduces a new permitting risk layer that is not yet priced into most models.

---

AI Drug Discovery Is Attracting Simultaneous Capital at Both the Company and Talent Layer

Two separate signals in this edition — Chai Discovery's $400M raise at $3.8B and departing OpenAI researcher Miles Wang's anticipated $200M raise for a new drug discovery startup — suggest that AI-driven drug design is moving from a niche theme to a capital-dense investment category at speed.

"Chai Discovery...develops AI models that help drugmakers design molecules by predicting and reprogramming how molecules interact, raised a $400 million round at a $3.8 billion post-money valuation."

"Miles Wang...is leaving the ChatGPT maker to launch a new startup focused on developing AI models for drug discovery...he may hit the ground running with a $200 million round of funding."

The convergence of top frontier AI talent and institutional capital into this single vertical in one newsletter cycle is a high-signal data point for investors tracking where applied AI value will accrue.