Frontier labs come for biology

---

1. Key Themes

Theme 1: Frontier AI Labs Are Moving Into Life Sciences, Disrupting the VC Thesis That Protected Biotech Startups

For years, biotech AI investors assumed frontier labs would stay out of specialized domains like drug discovery. OpenAI's launch of GPT-Rosalind has shattered that assumption.

"Life sciences have moved from being a side interest to a top priority for leading AI labs." — Simon Turner, Partner, Sofinnova Partners

"VCs have poured millions of dollars into AI life sciences startups under the assumption that their applications—like drug discovery—require proprietary data and specialized model architectures that frontier labs wouldn't bother to build. Although investors agree that this assumption is now out the window, they see the news as a positive."

Despite the disruption, investors remain bullish on specialized startups because proprietary data and clinical complexity create durable moats that generalist models cannot easily overcome.

"We are still always going to have gaps in what we need in the world, to actually test, develop and do more clinical research." — Alex Morgan, Partner, Khosla Ventures

---

Theme 2: Google DeepMind, Not OpenAI, Is the Real Threat to Watch in Biology AI

While OpenAI is grabbing headlines with GPT-Rosalind, analysts point to Google DeepMind as the more formidable competitor in biological AI — with a meaningful head start.

"DeepMind remains the competitor worth watching most closely, with a multi-year head start on domain-specific biological reasoning. Rosalind is OpenAI entering a race Google is already running." — Dimitri Zabelin, PitchBook Senior AI Analyst

This reframes the competitive narrative: OpenAI is the newcomer in this domain, not the pacesetter.

---

Theme 3: PE Is Rotating into "AI-Proof" Blue-Collar Trades — Garage Doors Are the New HVAC Rollup

Private equity is flooding into residential trades businesses as a deliberate hedge against AI disruption of white-collar services. Garage door installation and repair businesses are the latest rollup target, mirroring the highly successful HVAC consolidation playbook.

"The market has experienced a surge in competition, participants say, as PE firms that typically focused on other segments look to the trades for protection against the disruptive threat of AI."

"I think some of the historical perspective on blue collar services and just construction, as being kind of a dirty word in private equity, that's gone." — Paul Giovannoni, Partner, IFM

The scale of activity is significant: there were 29 PE-backed garage door deals in 2024 and 26 in 2025, "both years dwarfed what came before." The Oak Hill Capital acquisition of Guild Garage Group was valued at $800 million at a 16x EBITDA multiple — "more in line with an IT or data services business than a typical residential services company."

---

Theme 4: Agentic AI Adoption Is Real but Uneven — High-Stakes Decisions Remain Human

Agentic AI is gaining measurable traction in defined, lower-risk functional areas but hitting hard limits wherever judgment and accountability matter.

"Agentic AI adoption is making strides in IT, customer support and go-to-market functions, but remains more limited in areas requiring high-stakes judgment."

This signals a clear near-term opportunity: tools targeting IT ops, customer service automation, and sales/marketing workflows are in the adoption sweet spot. Autonomous decision-making in legal, medical, or financial domains faces a longer runway.

---

Theme 5: Private Capital Is Reshaping Sports Team Ownership at Record Valuations

The San Diego Padres are reportedly in talks to sell for ~$3.9 billion — fueled by PE money and setting a new high-water mark for MLB franchises. The broader trend of PE ownership of professional sports teams is accelerating.

"The San Diego Padres are in talks to sell to PE billionaire Jose E. Feliciano and his wife, Kwanza Jones, for around $3.9 billion."

This follows a broader pattern of private capital treating sports franchises as an alternative asset class with scarcity value, recurring revenue, and media rights upside.

---

2. Contrarian Perspectives

Contrarian 1: OpenAI Entering Life Sciences Is Good News for Biotech Startups, Not a Death Knell

The consensus fear would be that a frontier lab entering a niche domain kills the startups operating there. The article argues the opposite — VCs see OpenAI's move as a validation and a catalyst.

"Although investors agree that this assumption [that frontier labs wouldn't enter life sciences] is now out the window, they see the news as a positive."

The reasoning: GPT-Rosalind's current capabilities are limited to synthesis and planning, not invention. Proprietary pharma data remains a structural moat. As one expert put it: "If the drug does not exist yet, Chat-GPT cannot invent." — Duxin Sun, Associate Dean of Research, University of Michigan College of Pharmacy

The practical implication: startups with proprietary biological datasets and wet-lab integration remain defensible — and OpenAI's entry may accelerate investor interest in the entire sector.

---

Contrarian 2: The UK's Sovereign AI Fund Is Cosmetic, Not Strategic

While the UK government's sovereign AI fund is framed as a bold national initiative, the article's framing suggests it is grossly insufficient relative to the structural obstacles the UK faces.

"A new UK sovereign AI fund offers startups a boost, but funding constraints, high electricity prices and talent shortages limit British ambitions."

The fund is characterized as "a drop in the ocean" — signaling that government capital alone cannot compensate for systemic infrastructure deficits. For investors, this suggests UK AI startups face persistent headwinds that government support will not resolve in the near term.

---

Contrarian 3: Garage Door Businesses Are Trading at IT-Sector Multiples — A Valuation Warning Signal

The enthusiasm around garage door rollups may be creating a pricing bubble in a sector that, while attractive, carries real operational complexity.

"At a 16x EBITDA multiple, according to PitchBook data, pricing was more in line with an IT or data services business than a typical residential services company."

"Garage door investments are no gimme. One of the things that makes the sector attractive—fragmentation—also makes it difficult to find good targets, at least of the size and sophistication that PE firms typically seek in a platform company."

The flood of PE capital chasing the same fragmented market risks compressing returns just as valuations peak — the classic late-cycle rollup dynamic.

---

3. Companies Identified

---

4. People Identified

---

5. Operating Insights

Insight 1: Biotech Startups Should Lean Into Proprietary Data as the Primary Moat Against Frontier Lab Competition

The structural reason generalist models cannot displace specialized life sciences startups is not model architecture — it's data access. Founders in this space should prioritize acquiring, generating, and licensing proprietary experimental datasets, as this is what frontier labs cannot replicate quickly.

"Much of the crucial experimental data in the pharmaceutical sciences is proprietary and owned by large pharmaceutical laboratories."

Tactical implication: Partnerships with pharma companies for data access — not just compute or model capability — are the highest-leverage business development moves for biotech AI startups.

---

Insight 2: PE Operators Building Trade Services Rollups Should Prioritize Platform Quality Over Volume of Add-Ons

The garage door rollup wave is creating a crowded, competitive acquisition market. The article explicitly warns that fragmentation — while attractive for consolidation — makes it hard to find quality platform companies.

"One of the things that makes the sector attractive—fragmentation—also makes it difficult to find good targets, at least of the size and sophistication that PE firms typically seek in a platform company."

Tactical implication: Operators building in this space should compete on platform quality, management talent, and technology enablement (routing, scheduling, customer experience) rather than raw deal volume — and be disciplined on entry price as multiples climb toward IT-sector levels.

---

Insight 3: SPV Managers Should Benchmark Their Carry Structures — The Market Has Moved Below 20%

A Sydecar analysis of hundreds of SPVs found the traditional "2 and 20" is no longer the default. Only 25% of SPV deals used the traditional 20% carry, with both average and median carry coming in lower.

"SPV managers are adjusting fees and carry to match deal complexity, LP expectations, and the realities of running a lean team. Management fees are more common in later-stage deals, reflecting the additional diligence and sourcing work required."

Tactical implication: SPV managers using standard 20% carry without justification may be losing LP commitments to more competitively structured peers — especially in secondary transactions, where the article notes the "most divergence from standard terms."

---

6. Overlooked Insights

Overlooked Insight 1: European VC Is in a Prolonged Fundraising Crisis — Now at a Decade Low

Buried in the Chart of the Day is a significant structural data point that deserves more attention: European VC fund capital commitments have been declining since 2022 and hit their worst annual totals in a decade in 2025.

"Capital commitments to European VC funds have been on a downward slide since 2022, resulting in their worst annual totals in a decade in 2025."

For investors and founders operating in Europe, this means the capital environment is meaningfully tighter than headline AI enthusiasm might suggest — and the UK's AI fund being described as "a drop in the ocean" reinforces this structural gap. The recovery, the article hints, is uneven and not inclusive of all fund managers.

---

Overlooked Insight 2: 2018 Vintage North American VC Funds Have Barely Returned Capital Despite Decent IRR

The Daily Benchmark data reveals that 2018 vintage North American VC funds show a median DPI (Distributions to Paid-In capital) of just 0.15x — meaning LPs have received back less than 20 cents on the dollar in actual cash, despite a median IRR of 11.90%.

This is a quiet but important signal for LP portfolio construction: paper returns in VC remain largely unrealized even for funds now 7–8 years into their lifecycle, underscoring the ongoing liquidity crisis in private markets that is reshaping how LPs think about new commitments.