Short answer: pieces of it will land, but the full “moon‑shot” is unlikely to fly on the timeline the White House is selling. Innovation‑friendly deregulation can move fast; the infrastructure, workforce, and geopolitical pieces cannot. Expect quick wins around model deployment and limited export‑control tweaks, but years of trench warfare over energy, chips, and international standards.Below is a deeper score‑card so you can see where the big friction points lie and what would have to change for the plan to deliver.
1. What the plan gets right (and will probably accomplish)
| Area | Why it’s achievable | Real‑world tail‑winds |
|---|---|---|
| Removing federal red tape on AI R&D and procurement | Most of this is executive‑branch housekeeping—rescinding Biden‑era guidance, issuing new FAR clauses, and using the existing Chief AI Officer network. No new money or legislation needed. | Agencies already piloted hundreds of LLM use‑cases in 2024–25; they were waiting for a green light to scale. Expect uptake in back‑office tasks and public‑facing chatbots within 12‑18 months. |
| Open‑source / open‑weight push | The private sector wants this; procurement preferences plus NAIRR compute credits can tip the scales. Academic and VC communities are loudly in favor. | Meta’s Llama family and the coming Open‑Weight GPT variant give the policy an off‑the‑shelf success story. publicknowledge.org |
| Stronger national‑security evaluations via CAISI | Renaming the AI Safety Institute narrows its mission but gives Commerce a clear, funded mandate to test for CBRN and cyber risks. That plays well on the Hill. | Ongoing bipartisan bills (Chip Security Act, etc.) show Congress is ready to fund verification/enforcement tools. www.theverge.comwww.reuters.com |
| Exporting “full‑stack” U.S. AI to allies | Commerce can assemble consortia and bundle Ex‑Im/DFC financing without waiting for new statutes. | Many emerging‑market governments would rather buy U.S. tech than Chinese—if financing is turnkey. www.reuters.com |
2. Where the wheels are likely to come off
2.1 Energy & grid build‑out (Pillar II)
Demand reality: AI‑driven data‑centre load could grow 30‑fold by 2035 www.reuters.com and push capacity prices to record highs (PJM’s latest auction already jumped 22 %) www.ft.com.Bottlenecks
- Interconnection queues are averaging 3–5 years; NEPA shortcuts can shave months, not years.
- Long‑lead assets like nuclear SMRs and gas pipelines still face state siting fights and supply‑chain limits.
- Transmission is the real choke‑point; the plan talks permitting but not cost‑allocation or FERC authority—exactly the issues that sank past grid bills. Verdict: Expect lots of groundbreaking ceremonies, but meaningful new gigawatts won’t arrive before 2029–30. In the meantime, higher power prices will erode the “cheap energy dominance” narrative.
2.2 Semiconductor manufacturing
The plan leans on a “revamped” CHIPS Program Office but the administration has purged ~40 % of its staff and publicly trashed the law www.yahoo.comwww.datacenterdynamics.com. One marquee Michigan fab has already collapsed over funding uncertainty apnews.com.Verdict: Unless Congress restores predictable, multi‑year incentives, the fab spree slows and the U.S. stays reliant on TSMC/Samsung for cutting‑edge nodes.
2.3 Labor & skills
BLS projects strong growth for AI‑adjacent occupations (e.g., +18 % software devs 2023‑33) but also flags high displacement risk in clerical/legal support roles www.bls.govwww.bls.gov. The plan’s retraining ideas are good—but funding is thin and the same agencies face hiring freezes. Without money, the “AI Workforce Research Hub” is a press release.
2.4 Global rule‑setting
While the U.S. deregs, Brussels is implementing the EU AI Act and pushing OECD/G7 codes of conduct. Allies may balk at adopting U.S. tech if Washington looks cavalier on privacy or bias. The plan’s call to “counter Chinese influence” is sound, but it offers carrots (export packages) without addressing allied concerns about rights and safety.
3. Strategic contradictions to watch
| Tension | Why it matters |
|---|---|
| Open‑source vs. export controls | You can’t both flood the world with American open‑weights and guarantee adversaries don’t fine‑tune them. Location‑locked chips don’t stop model‑weight leaks. |
| Free‑speech absolutism vs. disinformation | Stripping “misinformation” out of the NIST Risk Framework may please culture‑war allies, but it undercuts trust for health‑care, finance, and election applications that regulators still have to police. |
| Deregulate now, regulate later | History (railroads, radio, the Internet) shows that heavy deregulation often locks in incumbents—the opposite of the plan’s anti‑“Big Tech” rhetoric. |
4. Bottom‑line scenarios
| Through 2027 | 2028‑35 outlook |
|---|---|
| Most likely (60 %): Innovation boom in software & services, grid stress intensifies, chip fabs lag. GDP bump of ~0.3–0.5 ppt, but electricity prices up, and allies adopt a patchwork of EU/US standards. | Growth tapers as energy costs bite; U.S. retains model leadership but cedes some infra edge to nations with cheaper power (e.g., Canada, Nordics, Gulf states). |
| High‑success (20 %): Congress funds transmission & NAIRR, CHIPS 2.0 passed, bipartisan export‑control enforcement. U.S. secures both model and hardware leadership. | Sustained 1 ppt AI contribution to GDP, relative energy costs stabilize as nuclear and geothermal scale. |
| Low‑success (20 %): Legal challenges stall permitting reforms, budget hawks kill spending, Europe and Japan coalesce around stricter safety standards. | U.S. remains tops in frontier R&D but loses manufacturing & infra race; AI gains concentrated in a few coastal metros, widening inequality. |
5. How to raise the odds of success
- Treat transmission as critical defense infrastructure—use Defense Production Act Title III plus FERC backstop siting to build multi‑state HVDC corridors.
- Keep the CHIPS guardrails but streamline paperwork instead of scrapping the act; certainty beats ideology for private capital.
- Fund demand‑side R&D (efficient architectures, optical compute) to blunt the power crunch rather than betting solely on new generation.
- Layer in mandatory transparency metrics (nutrition‑label‑style) rather than removing DEI & misinformation language outright; that reassures allies without heavy licensing.
- Swap “blacklist and pray” export controls for a club‑goods approach—offer allies privileged access to U.S. models/chips in exchange for matching controls and data‑sharing.