Leaked draft rules and Commerce messaging point away from Biden's global diffusion framework and toward discretionary, country-by-country deals tied to U.S. leverage. My base case is a tougher but more negotiable regime: more licensing, more bilateral bargaining, and more demands that buyers anchor compute or capital in the U.S. ([axios.com](https://www.axios.com/2026/03/05/trump-ai-chip-clash-white-house))
Verdict: The direction is clearer than the final text. Axios reported a White House fight over a BIS draft, Tom's Hardware cited Commerce saying it would not revive the old diffusion rule, and BIS already moved in January to case-by-case licensing for certain China chip exports (Axios, 2026-03-05; Tom's Hardware, 2026-03-06; BIS, 2026-01-13). Expect a bargaining-heavy system that mixes security review with industrial policy. ([axios.com](https://www.axios.com/2026/03/05/trump-ai-chip-clash-white-house))
Washington builds a clear allied framework with transparent thresholds and fast approvals for trusted buyers. That preserves U.S. security goals while keeping American firms commercially competitive. Compliance costs rise, but predictability improves enough for long-range planning.
The U.S. replaces universal tiers with discretionary bilateral bargaining. Large sales increasingly require political assurances, monitoring, and domestic investment sweeteners. Firms can still sell abroad, but policy risk becomes a permanent line item.
Rules become opaque, allies feel downgraded, and rivals accelerate alternatives. Export licensing slows deals enough to shift demand toward non-U.S. stacks and local fabs. Security goals partly hold, but U.S. firms lose influence in fast-growing markets.
Model efficiency or non-U.S. hardware improves so quickly that chip-centric controls lose bite. Governments then shift attention from chips to cloud access, energy access, and model weights. The policy fight broadens from semiconductors to the whole compute chain.
Developments: The administration likely favors negotiated country packages over a fully standardized global rule. Big buyers seek explicit political backing before ordering large clusters. Compliance teams gain influence inside chipmakers and cloud firms.
Risks: Draft language may change abruptly and disrupt sales cycles. Allies could react badly if approvals feel arbitrary. China and other rivals could use the uncertainty to deepen non-U.S. supply chains.
Outlook: Over the next year, discretion matters more than doctrine. Sales remain possible, but only with more government touchpoints. Policy noise stays high.
Developments: Export compliance shifts from back-office screening to front-line commercial strategy. Governments bundle chips with cloud, security, and investment commitments. Sovereign AI programs increasingly negotiate directly with Washington and U.S. vendors.
Risks: Smaller firms may struggle to navigate the cost and complexity. Political changes could rewrite country priorities midstream. Excess discretion may invite lobbying and accusations of favoritism.
Outlook: Two years out, the winners are firms that can sell policy certainty alongside hardware. Pure product competition becomes less decisive. The market gets more political.
Developments: Trusted partners push for standing privileges, shared monitoring, and reciprocal industrial access. More export deals include audit rights, data-center conditions, and cybersecurity clauses. Buyers seek to lock in supply before the next rule shift.
Risks: Reciprocity requests may collide with domestic industrial goals. Governments can overbuild compute if diplomatic prestige outruns real demand. A new crisis could cause sudden blanket tightening.
Outlook: Three years out, the regime is likely thicker with treaties, side letters, and exceptions. Formal law matters, but implementation matters more. The center of gravity stays bilateral.
Developments: Compute ecosystems cluster into U.S.-aligned, China-aligned, and mixed-access zones. Chips, cloud, power, networking, and safety rules are packaged together. Frontier buyers diversify suppliers to reduce exposure to any one policy center.
Risks: Regionalization can reduce scale efficiency and slow diffusion of useful technology. Cross-border research may become harder. Firms could face conflicting legal duties across jurisdictions.
Outlook: Five years out, export control is no longer a narrow trade issue. It is part of a regional industrial architecture. Strategic alignment shapes market access.
Developments: By the mid-2030s, compute access is likely treated like energy security or telecom sovereignty. Governments maintain reserve capacity, preferred vendor lists, and crisis protocols. National security reviews extend beyond chips to training data, orchestration software, and energy use.
Risks: Bloc hardening may reduce innovation spillovers. Escalation could produce tit-for-tat controls on tools, materials, or talent. Customers may pay much more for resilience than for raw performance.
Outlook: Ten years out, frontier compute is plausibly governed like strategic infrastructure. Open global markets persist at the edge, not the frontier. Policy durability matters as much as technical leadership.
Developments: Compute diplomacy broadens into power, cooling, land use, and cross-border cloud governance. Nations negotiate access to both chips and the physical systems that make large-scale training possible. Export policy becomes one module inside a larger strategic-tech treaty system.
Risks: A complex treaty web can entrench incumbents and shut out developing markets. Energy shortages or climate stress could make compute access more political still. Enforcement may become uneven and fuel gray markets.
Outlook: Twenty years out, the strategic object is not the chip alone. It is the whole compute stack. Countries that coordinate across trade, energy, and security gain the advantage.
Developments: Over half a century, the most sensitive controls likely move from specific chips to integrated systems, model capabilities, and critical energy networks. Hardware generations change too quickly for static lists to dominate policy. States focus on governing concentration points where power, data, software, and trust intersect.
Risks: Long-run overcontrol could fragment science and lock in inequality. Undercontrol could let destructive capabilities diffuse too easily. Institutional lag will always tempt governments to regulate yesterday's bottleneck.
Outlook: Fifty years out, export control survives but its target evolves. The core issue remains who can marshal frontier compute at scale. Bargaining power will belong to states and firms that control multiple bottlenecks at once.