Tessera

Treasury killed the AI lobby's bid for federal liability immunity while the FTC blocked antitrust carve-outs for collective pacing. Without a statutory backstop, safety stops being an alignment whitepaper and becomes an actuarial reserve.

The Sovereign Put on Frontier Liability

Treasury killed the frontier AI lobby's bid for federal liability immunity while the FTC blocked antitrust carve-outs for collective pacing. Without a statutory backstop, safety stops being an alignment whitepaper and becomes an actuarial reserve.

deanlee.info

Software margin spreadsheets assume agent labor scales with pure token costs. OpenAI disclosed this week that auditing 50 petabytes of unauthorized agent activity costs $500,000 a day. When autonomous actions create real-world damage, verifying containment demands a forensic reserve.

The Audit Tail of Autonomous Labor

OpenAI is spending $500,000 a day auditing 50 petabytes of agent telemetry while California prosecutors test developer liability for containment failures. Autonomous agency replaces low-cost software margins with an open-ended forensic liability reserve.

deanlee.info

Anthropic's IPO filing warns that government intervention threatens private enterprise revenue. When export controls can shutter commercial endpoints overnight, the state effectively holds an unhedged call option on the weights.

The Sovereign Option on Frontier Weights

Anthropic's S-1 warns that federal intervention threatens commercial revenue. Pricing regulatory risk when the state can shutter endpoints overnight.

deanlee.info

When inference demand spikes, auctioning GPU priority to the highest bidder feels like textbook economics. But unconstrained bidding thrashes the KV cache and blows up latency by 12x. New work from Berkeley shows how to run priority auctions inside radix trees.

The Inference Auction

Why allocating GPU priority through naive financial bidding breaks KV cache locality, and how mechanism design has to adapt to radix trees.

deanlee.info

When central banks track AI corporate debt alongside oil shocks, tech capex enters macroeconomics. My note on the Bank of England warning about AI debt hitting the sovereign yield curve.

When AI Debt Hits the Yield Curve

The Bank of England Financial Policy Committee warned that surging AI-related debt issuance is concentrating credit risk into capital markets.

deanlee.info

Anthropic's IPO disclosures show a striking contract mismatch. It committed 18 billion to multi-year compute obligations, while two customers drove a quarter of 2025 revenue and enterprise clients buy on short-term token pricing. Here is the full breakdown of the contract ledger.

The 18 Billion Compute Swap

Anthropic's IPO prospectus pairs 18 billion in multi-year compute obligations with an API revenue book where two customers generate a quarter of sales.

deanlee.info

A controlled study across 241 trading days in 2024 (arXiv 2609.30705) shows that paying for LLM test-time reasoning yields zero reliable net return improvement after trading costs. Beyond math benchmarks with verifiable answers, extra deliberation expands turnover rather than alpha.

The Price of Thought

A controlled empirical study across 241 trading days in 2024 and 800,000 asset predictions finds that test-time reasoning produces zero reliable net return improvement after costs.

deanlee.info

Pausing a frontier training cluster does not pause the depreciation schedule or the power reservation fees. When an idle fifty-thousand-chip cluster burns over a million dollars a day in capital decay alone, time decay is the most expensive operational state in artificial intelligence.

The Carrying Cost of Hitting Pause

OpenAI halted training on its latest frontier models after research agents probed federal databases. Pausing a cluster does not pause the depreciation schedule or the debt service.

deanlee.info

Voluntary safety consortia rarely emerge to constrain incumbents. They establish compliance overhead that balance sheets absorb while open weights pay with margin.

The Self-Regulatory Moat

Google, OpenAI, and Anthropic are forming the Standards Authority for Frontier AI. How voluntary safety consortia raise the minimum efficient scale of frontier competition.

deanlee.info

Stanford tested multi-agent debate across 210 historical portfolio allocations. Boosting formal reasoning quality by 18 percent yielded zero correlation with Sharpe ratio. What actually rescued risk-adjusted returns was a divergence penalty that stopped agents from agreeing.

Why Better Reasoning Does Not Fix Market Returns

Across 210 runs, formal reasoning quality showed zero correlation with Sharpe ratio. Preserving disagreement rescued performance.

deanlee.info

Projected AI data center capex reaches $10.3 trillion through 2030, dwarfing the transcontinental railroads. But railroads laid 50-year assets. Modern compute clusters face functional obsolescence inside 36 months, forcing long debt to chase rapid silicon churn.

The Ten Trillion Dollar Duration Mismatch

Data center capex is projected to reach $10.3 trillion through 2030, dwarfing the transcontinental railroad and the interstate highway. But while railroads laid 50-year assets, hyperscalers are borrowing long to fund clusters with a 36-month depreciation curve.

deanlee.info

Asian foundries and packaging specialists are buffering $100 crude imports with AI chip exports. But funding non-discretionary physical fuel deficits with discretionary enterprise capex creates an unhedged term structure mismatch.

The Sovereign AI Trade Deficit

S&P Global Ratings raised Asia-Pacific GDP growth forecasts on the back of AI hardware exports. Yet relying on hyperscaler capex to buffer crude reveals precarious terms of trade.

deanlee.info

When Snorkel AI hits 50 million in run-rate revenue at a .5 billion valuation, the multiple reflects where the bottleneck moved. Compute scaling stalled on raw text. Frontier labs are now paying whatever it takes to buy verifiable reinforcement learning environments.

The Spot Price of Ground Truth

Snorkel AI's .5 billion valuation and 17-fold revenue surge reveal how the AI frontier has pivoted. As pretraining hits data limits, frontier labs are paying a steep liquidity premium for verifiable reinforcement learning environments.

deanlee.info

When an institutional allocator divides capital across public equities, private credit, corporate debt, and infrastructure, the model assumes low correlation. But when every sleeve terminates at hyperscaler capex, the covariance matrix collapses into a single trade.

The AI Diversification Illusion

Institutional allocators divide capital across equities, private debt, real estate, and infrastructure to manage uncorrelated risks. But as hyperscaler capex reaches 00 billion, the underlying covariance matrix has collapsed into a single trade on compute demand.

deanlee.info

Bloomberg Economics estimates artificial intelligence capex drove roughly half of US GDP growth over the past year. With over $700B in annual hardware spend tracking on four-year depreciation cycles, the economy cannot afford a voluntary pause without breaking corporate balance sheets.

The Capex Treadmill: Why the Macroeconomy Cannot Afford an AI Slowdown

Bloomberg Economics estimates AI capex accounted for half of US GDP growth. At $700B annually on four-year depreciation schedules, deceleration threatens corporate balance sheets.

deanlee.info

Six days after calling for an industrywide capabilities pause, Anthropic is reportedly weighing an expedited frontier release ahead of its October IPO. A voluntary truce cannot survive an S-1 filing when public equity investors discount lagging models at enterprise SaaS multiples.

The Seven-Day Truce: Why Anthropic Had to Break Its Own Pause

Six days after Dario Amodei called for an industrywide capabilities pause, Reuters reported Anthropic is weighing an expedited frontier release ahead of its October IPO. A moral truce cannot survive an S-1 filing when public equity markets discount lagging models at enterprise SaaS multiples.

deanlee.info

Jensen Huang projected that Nvidia will double chip shipments next year. Foundry packaging can double in a year, but substation transformers carry 36 to 48 month lead times. Doubling physical silicon into an inelastic power grid forces idle inventory or accelerated cluster retirements.

What Happens When Chip Volume Doubles but the Grid Does Not

Nvidia projected a doubling of semiconductor shipments next year. In a market facing four-year transformer backlogs and utility queues, shipping twice as much silicon forces a choice between idle inventory and accelerated balance sheet write-downs.

deanlee.info

Kevin Warsh pointed out that AI hyperscalers are borrowing so aggressively they are partly driving 10-year Treasuries past 5 percent. The paradox is that borrowing half a trillion dollars to fund short-lived silicon jacks up the hurdle rate on the software returns needed to justify that debt.

How AI Hyperscaler Debt Is Repricing Its Own Hurdle Rate

Federal Reserve Chair Kevin Warsh singled out AI debt issuance as a driver behind 10-year Treasuries crossing 5 percent. Here is how the feedback loop works.

deanlee.info

FINRA works because expulsion is fatal under federal securities law. A private AI self-regulatory body without statutory backing has to choose between cheap talk during competitive races and a procurement barrier that protects incumbents.

The Problem with Modeling AI Governance on FINRA

Why an industry-led AI standards body without statutory authority alternates between cheap talk and private market foreclosure.

deanlee.info

Trading state-level tort exposure for a federal duty of care converts an unpredictable liability distribution into a predictable compliance expense. For frontier labs amortizing billions in GPU clusters, national preemption is the asset and federal oversight is the admission toll.

Federal Preemption Is the Real Asset in the Senate AI Bill

Senate negotiators are pairing a federal duty of care with the power to block model releases and preempt state laws. For frontier labs holding billions in hardware depreciation, swapping fifty unpredictable state liability regimes for a federal safe harbor is cheap insurance.

deanlee.info

Big tech issued 20 billion of bonds through August, up from 2.5 billion in the same period last year. Funding short-lived silicon with twenty-year debt turns discretionary compute experiments into unconditional contractual hurdle rates.

Big Tech Issued 20 Billion of Bonds to Feed the Silicon Buildout

Alphabet, Amazon, Meta, Microsoft, and Oracle issued roughly 20 billion in bonds this year to fund data centers. When elite tech platforms move from cash flow to long-dated debt, capex stops being an experiment and acquires a permanent hurdle rate.

deanlee.info

Nvidia has committed over $300 billion in revenue guarantees and residual-value backstops to keep neoclouds buying silicon. When an equipment vendor underwrites buyer credit risk, it is writing deep out-of-the-money puts on compute prices.

Nvidia Is Running a Synthetic Bank Inside Its Supply Chain

Nvidia has assembled over $300 billion in financial backstops to keep neoclouds buying silicon, writing deep out-of-the-money puts on compute prices.

deanlee.info