Rosa Del Mar

Daily Brief

Issue 43 2026-02-12

Crypto And Ai Policy Uncertainty As An Operational Constraint

Issue 43 Edition 2026-02-12 10 min read
General
Sources: 1 • Confidence: Medium • Updated: 2026-02-12 18:41

Key takeaways

  • Copyright treatment for AI training—whether models may learn from copyrighted works without reproducing them—is a key upcoming policy issue.
  • Goldman Sachs was the largest wholesale funder in the world 10 years ago and has since prioritized moving away from reliance on wholesale funding toward more stable sources.
  • Goldman spent about $6B on technology last year but could not spend $8B without reducing returns, implying efficiency savings are needed to increase investment while maintaining performance.
  • Geopolitics is tougher as the world shifts back toward multipolarity, increasing the risk of a geopolitical shock that slows growth compared to the post–Cold War era.
  • Ongoing uncertainty and aggressiveness from the FTC, including toward smaller tech deals, could shift M&A toward IP-style transactions rather than traditional acquisitions.

Sections

Crypto And Ai Policy Uncertainty As An Operational Constraint

  • Copyright treatment for AI training—whether models may learn from copyrighted works without reproducing them—is a key upcoming policy issue.
  • The 'Clarity Act' is a pending crypto market-structure bill intended to define how different token types are classified under rules.
  • Banning AI or restricting underlying mathematics would cause the U.S. to lose the AI race to China with long-term strategic consequences.
  • Parts of the U.S. crypto industry were effectively banned by a prior U.S. administration via executive pressure rather than legislation or formal legal process, including the use of debanking tactics.
  • A prior U.S. administration treated essentially all tokens and some NFTs as securities via enforcement actions described as extreme.
  • A patchwork of 50 state-level AI laws would make it effectively impossible for new companies to comply and innovate.

Goldman Governance, Scale, And Funding Resilience

  • Goldman Sachs was the largest wholesale funder in the world 10 years ago and has since prioritized moving away from reliance on wholesale funding toward more stable sources.
  • Goldman Sachs shifted from having zero deposits 15 years ago to about $500B in total deposits, including a digital deposit platform with over $200B, funding roughly 40% of the firm with deposits.
  • Goldman Sachs has maintained a partnership-like culture post-IPO, with roughly 450 partners whose compensation is correlated to overall enterprise performance.
  • In mature financial services businesses, scale provides leverage and latitude during turbulence, making scale a central long-term strategic requirement for Goldman Sachs.
  • Goldman Sachs leadership concluded that operating as a public company requires top-down strategic direction to make the organization’s parts add up to more than their sum.
  • Goldman Sachs believes it must continue increasing balance-sheet scale over the next 5–15 years because it is currently far smaller than JPMorgan and organic scale-building is difficult in mature businesses.

Enterprise Ai Transformation: Tooling Distribution Vs Process Reengineering Under Regulatory Gating

  • Goldman spent about $6B on technology last year but could not spend $8B without reducing returns, implying efficiency savings are needed to increase investment while maintaining performance.
  • Goldman launched '1GS 3.0' to reimagine six firm processes; the capacity impact is described as very significant but not publicly quantified.
  • Goldman’s first AI focus is distributing tools/models broadly to employees so they can experiment and discover productivity gains in client work.
  • The more consequential AI opportunity is reimagining core enterprise operating processes for automation and efficiency, then reinvesting savings into growth areas.
  • Large-scale process reimagination is difficult because it threatens existing organizational 'empires' and therefore must be driven top-down.
  • Regulatory clearance requirements significantly slow Goldman’s ability to deploy AI tools compared with companies that can deploy without such gating.

Macro Drivers And Concentration Of Growth Inputs

  • Geopolitics is tougher as the world shifts back toward multipolarity, increasing the risk of a geopolitical shock that slows growth compared to the post–Cold War era.
  • A combination of fiscal stimulus, a rate-cutting cycle, a capital investment supercycle, and a deregulatory unwind makes the U.S. economy hard to slow.
  • Last year, the four largest companies contributed about 1% to U.S. GDP growth through roughly $400B of spending.
  • The current U.S. macro setup for investable and financial assets is the best 'sweet spot' seen in decades despite substantial global complexity.

Capital Markets Activity As A Function Of Regulatory Confidence

  • Ongoing uncertainty and aggressiveness from the FTC, including toward smaller tech deals, could shift M&A toward IP-style transactions rather than traditional acquisitions.
  • Capital markets activity is tied to confidence, and a tough regulatory environment suppresses confidence and therefore M&A, IPOs, and capital raising.
  • Deal activity is expected to pick up significantly, and this year could be the biggest M&A year in history with a larger IPO year as confidence improves.

Watchlist

  • Geopolitics is tougher as the world shifts back toward multipolarity, increasing the risk of a geopolitical shock that slows growth compared to the post–Cold War era.
  • Ongoing uncertainty and aggressiveness from the FTC, including toward smaller tech deals, could shift M&A toward IP-style transactions rather than traditional acquisitions.
  • The 'Clarity Act' is a pending crypto market-structure bill intended to define how different token types are classified under rules.
  • Banning AI or restricting underlying mathematics would cause the U.S. to lose the AI race to China with long-term strategic consequences.
  • Copyright treatment for AI training—whether models may learn from copyrighted works without reproducing them—is a key upcoming policy issue.
  • Models trained on widely available information may not be able to produce differentiated investment outperformance.

Unknowns

  • What are the current levels and trends of Goldman’s wholesale funding reliance versus deposit funding, including stability and cost of funds in stress scenarios?
  • What specific six processes are included in '1GS 3.0', and what quantified outcomes (capacity, cycle time, error rates, headcount redeployment) have been achieved?
  • What evidence supports the claim that some AI-era companies reached $100M to $1B+ in revenue in under a year, and how common is this across cohorts?
  • Is the claim that Andreessen Horowitz raised about 18.3% of all U.S. venture capital in 2025 accurate, and what definition of 'U.S. venture capital raised' is being used?
  • What specific regulatory actions or data substantiate or refute claims of debanking-driven crypto suppression and unusually broad securities classification of tokens/NFTs?

Investor overlay

Read-throughs

  • Sustained crypto market-structure uncertainty and discretionary enforcement could keep capital formation cautious and shift industry behavior toward compliance-heavy models, while pending bills like the Clarity Act become key catalysts for confidence and activity.
  • AI regulation may tilt toward application-focused rules, but unresolved copyright treatment for training data could become the binding constraint for AI product rollouts, licensing economics, and litigation exposure across model developers and content owners.
  • Aggressive FTC posture, including toward smaller tech deals, could re-route M&A from full acquisitions toward IP or asset-style transactions, making regulatory confidence a primary driver of deal volume and deal structure.

What would confirm

  • Clear legislative or regulatory movement on token classification and market structure that reduces discretionary enforcement, alongside a measurable pickup in crypto-related capital markets activity tied to improved confidence.
  • Court rulings, legislation, or widely adopted industry licensing frameworks that clarify whether AI models may train on copyrighted works without reproducing them, and observable changes in deployment pace for regulated enterprise AI use cases.
  • Publicly observable deal structures increasingly using IP, asset purchases, or carve-outs in situations where traditional acquisitions face FTC friction, along with broad-based increases in announced deals as confidence improves.

What would kill

  • Regulatory actions that increase uncertainty for tokens or expand enforcement ambiguity, accompanied by continued muted issuance, listings, or investment activity despite improving market conditions.
  • Legal outcomes or policy that materially restrict training on copyrighted works without scalable licensing alternatives, leading to delayed releases or curtailed model training plans rather than continued application-led progress.
  • FTC posture materially eases without a corresponding change in deal structures or activity, or deal confidence improves while M&A remains constrained for reasons unrelated to regulation in the summary.

Sources