🧬 Virtual cell team-up
1. Key Themes
Theme: Biology is the next AI frontier, and the bottleneck is data, not algorithms
AI for biology depends on generating data that doesn't exist yet
Unlike language or image AI, which could train on existing internet-scale corpora, biology requires new physical measurement.
"Biology presents an additional challenge: Much of the information AI needs doesn't exist yet and has to be painstakingly measured from the physical world."
The "universal virtual cell" is a coalition play
Public and private actors are pooling resources to create and standardize data.
"Biohub, the Department of Energy, the National Institutes of Health, Google DeepMind, Isomorphic Labs, Meta and a collection of scientific organizations are collaborating to create and standardize data for what Biohub calls a 'universal virtual cell.'"
The value proposition: virtual experiments that make lab work more valuable
The goal is triage, using AI to decide which expensive wet-lab experiments are worth running.
"If we can put more and more reasoning and intelligence into every single question that we actually ask in the lab, the value of those empirical results will be far greater."
Long-term payoff is personalized disease modeling
"Further out, Rives envisions models that could examine an individual's disease and predict its molecular causes and the best way to intervene."
Theme: AI compute is scarce on paper but underutilized in practice
Utilization is strikingly low even amid a shortage
Idle capacity coexists with researchers who can't get access.
"Independent, single-tenant data centers average less than 15% net computing utilization, leaving expensive capacity unused even as researchers face shortages, according to a paper the consortium produced with the announcement."
Pooling capacity could change the economics of the AI buildout
"Building AI infrastructure has become a multitrillion-dollar bet largely because of the high cost of accessing scarce computing resources."
"The creators of the compute grid want to make it easier for idle computing capacity to be used, a goal that could significantly alter the direction of the AI boom if it were to take off."
Open standards and geopolitics frame the pitch
"The best way to scale AI in America efficiently, and stay at the frontier and stay competitive with China, is to be" coordinated around an open standard.
Theme: Power demand is the physical constraint behind the AI buildout
Data center power demand is projected to surge
"Goldman Sachs Research projects AI will drive an increase of 170% in global data center power demand by 2030, adding the equivalent of Japan's entire electricity consumption."
Hyperscalers are securing nuclear supply
"Google and Constellation Energy struck a nuclear energy deal."
Theme: Consumer AI safety scrutiny is intensifying
Independent testing challenges teen safeguards
"Common Sense Media said ChatGPT missed more than 1 in 4 instances in which its testers determined a crisis referral was warranted."
Open-source AI misuse is a growing risk area
"A harrowing and important read from Bloomberg on how child predators are using open-source AI tools to make unlimited illegal images."
2. Contrarian Perspectives
Scaling laws will hold in biology, which many assumed was too messy for them
The consensus worry was that cellular biology might not follow the predictable data-to-performance curves seen in language models. Rives, who flagged this as a major open question in April, now argues the opposite.
"It's worked in every field, and it works in biology too," Rives said, pointing to AI's progress in protein biology.
If true, this implies that spending on data generation has predictable returns, a thesis that favors well-funded, data-generating consortia over clever-algorithm startups.
The compute shortage may be partly a coordination failure, not a pure supply problem
The dominant narrative is that chips are scarce and demand is insatiable, which justifies multitrillion-dollar buildouts. The compute grid argues much of the problem is fragmentation and idle capacity.
"Independent, single-tenant data centers average less than 15% net computing utilization."
The implication is that a market-structure fix (pooling, open standards) could relieve scarcity without proportional new capex, which is a potential headwind to the "build everything" thesis.
Competitors can be incentivized to cooperate on open data through temporary exclusivity
Instead of treating proprietary data as the moat, Biohub trades a time-limited advantage for broad participation.
"We have to have some incentive for commercial players to be a part of this, and the embargo period creates that." The commercial partners "will have one year of exclusive access to the data they develop before it is shared publicly."
3. Companies Identified
Biohub (Chan Zuckerberg Biohub)
- Description: Mark Zuckerberg-backed science organization leading the virtual cell effort.
- Why mentioned: Convening the coalition to generate and standardize data for a "universal virtual cell."
- Quotes: "Mark Zuckerberg's Biohub is partnering with Google and the federal government in its ambitious effort to use AI for generating vast quantities of biological data that can predict how cells behave." Also: "When Biohub announced its initial $500 million effort in April..."
- Description: Google's AI research lab.
- Why mentioned: Collaborator in the virtual cell effort.
- Quotes: "Biohub, the Department of Energy, the National Institutes of Health, Google DeepMind, Isomorphic Labs, Meta and a collection of scientific organizations are collaborating..."
Isomorphic Labs
- Description: AI drug discovery company.
- Why mentioned: Named commercial/scientific partner in the virtual cell consortium.
- Quotes: Same quote as above.
Meta
- Description: Zuckerberg's social/AI company.
- Why mentioned: Named partner in the virtual cell collaboration.
- Quotes: Same quote as above.
National Compute Grid (coalition)
- Description: A coalition of AI startups, cloud providers, researchers and investors pooling AI compute.
- Why mentioned: Attempting to address the chip supply crunch by unlocking idle capacity.
- Quotes: "A coalition of AI startups, cloud providers, researchers and investors is launching a National Compute Grid, an effort to pool AI computing to help address a supply crunch."
- Description: Maker of ChatGPT.
- Why mentioned: Its teen ChatGPT experience was criticized as unsafe; it disputes the findings.
- Quotes: "We welcome rigorous independent evaluation, but we do not believe Common Sense Media's testing accurately reflects how ChatGPT's teen safeguards work in practice."
Common Sense Media (Youth AI Safety Institute)
- Description: Nonprofit that rates media and technology for kids.
- Why mentioned: Published testing of 4,000+ prompts finding gaps in ChatGPT's teen safeguards.
- Quotes: "Common Sense Media is urging OpenAI to keep teens off ChatGPT, saying its new teen experience poses an 'unacceptable risk.'"
Google and Constellation Energy
- Description: Tech giant and nuclear power operator.
- Why mentioned: Struck a nuclear energy deal, signaling hyperscaler power procurement.
- Quotes: "Google and Constellation Energy struck a nuclear energy deal."
- Description: Investment bank and newsletter sponsor.
- Why mentioned: Research projecting data center power demand growth.
- Quotes: "Goldman Sachs Research projects AI will drive an increase of 170% in global data center power demand by 2030."
4. People Identified
- Description: Head of science at Biohub.
- Why mentioned: Primary architect and spokesperson for the virtual cell vision.
- Quotes: "We're at the beginning of a new scientific paradigm with AI." Also: "The big challenge in biology is to bridge that gap between compute and the digital world and the real physical world of biology and life. The way to do that is through data."
Mark Zuckerberg
- Description: Meta CEO and Biohub backer.
- Why mentioned: His Biohub is leading the effort.
- Quotes: "Mark Zuckerberg's Biohub is partnering with Google and the federal government..."
Anjney Midha
- Description: A leader of the National Compute Grid effort.
- Why mentioned: Articulated the open-standard, competitiveness-with-China rationale.
- Quotes: "The best way to scale AI in America efficiently, and stay at the frontier and stay competitive with China, is to be" coordinated around an open standard.
Tom Siegel
- Description: Head of Common Sense Media's Youth AI Safety Institute.
- Why mentioned: Delivered the sharpest critique of ChatGPT's teen safety.
- Quotes: "ChatGPT is not safe for kids to use."
5. Operating Insights
Design incentives that make competitors fund shared infrastructure
The embargo model gives paying participants early access while ensuring eventual openness. This is a replicable playbook for any consortium that needs both private money and public legitimacy.
"The embargo period creates that" incentive; commercial partners get "one year of exclusive access to the data they develop before it is shared publicly."
Invest in data generation, not just models, in domains where data is the constraint
In fields where data doesn't pre-exist, owning or shaping data pipelines can matter more than model architecture.
"The way to do that is through data."
Look for utilization arbitrage in capital-intensive infrastructure
If single-tenant assets run under 15% utilization, there is an opportunity for brokerage, scheduling, and pooling layers that monetize idle capacity or give startups cheaper access.
"Independent, single-tenant data centers average less than 15% net computing utilization."
6. Overlooked Insights
The compute grid is explicitly opening access to the public sector
Government, education and national lab users are named as early beneficiaries, which could position the grid as a quasi-public utility and a channel for government demand.
"The paper says the grid is opening access to public-sector employees and teams, including government, education and national laboratory users."
Parental-alert gaps may partly be a linking-latency issue, which exposes a product design lesson
OpenAI says safeguards weren't active during testing because account linking takes hours, suggesting a real vulnerability window for new parent-teen setups even if one disputes the severity.
"OpenAI says much of Common Sense's testing of parental alerts occurred before the parent and teen accounts had finished linking, a process the company says can take several hours."