US deregulation accelerates AI market growth

President Donald Trump’s administration has rolled out a 90-policy Artificial Intelligence Action Plan that dismantles the Biden-era guardrails, re-casts U.S. dominance as national policy, and sets off a global scramble to catch up. The blueprint, released on July 23, promises lighter regulation, faster commercial adoption, and an export drive designed to make American AI the world’s default standard.

The move ends months of suspense in Washington and Silicon Valley over how the White House would redirect federal tech strategy — and it delivers in sweeping fashion.

A Dramatic Pivot From the Biden Playbook

The administration calls its roadmap “The Trump administration’s 90+ policy AI Action Plan” — a direct rebuttal to the Biden team’s risk‑centric stance. At its core, the White House declares that innovation, not precaution, will keep America ahead of China.

Three Executive Orders, One Agenda

To lock the change in, the president signed three directives:

• Executive Order 14179 “Removing Barriers to American Leadership in Artificial Intelligence.” The January order “serves as the foundational directive, establishing U.S. global AI dominance as national policy.”
• The July “Preventing Woke AI in the Federal Government” decree forbids agencies from buying systems deemed ideologically biased. Contractors must ensure their tools are “objective and free from top‑down ideological bias,” or risk losing the deal.
• A third order “promotes American AI technology exports through full‑stack packages” — chips, models, and code — to allies. A companion measure to “streamline data‑center infrastructure” drew applause from hyperscalers but lacks detail. 

Regulation Repealed, Institutes Disbanded

The plan “directly targets what the administration characterizes as Biden’s ‘dangerous’ AI policies, rescinding Executive Order 14110,” the 2023 mandate that required safety disclosures and ethics tests. It also “eliminates requirements for AI companies to report safety testing results to the government and dissolves the U.S. AI Safety Institute’s regulatory oversight role.”

Diversity, climate, and misinformation checks are removed as “unnecessary barriers” under the new plan, which “removes diversity, equity, and inclusion considerations from the NIST AI Risk Management Framework.”

Chip Export Rules: Crackdown and Reversal

Competition with Beijing looms large. In April, the White House “significantly expanded AI chip export restrictions to China, barring companies like Nvidia and AMD from selling their H20 and MI308 processors.” Yet under industry pressure, the administration “reversed course in July, allowing Nvidia and AMD to resume AI chip sales to China.” Officials also “rescinded Biden’s AI Diffusion Rule in May 2025, replacing it with undefined future regulations.”

Critics say the zig‑zag leaves exporters guessing — and China strategists cheering. New license requirements for once‑compliant chips add another layer of uncertainty.

Power‑Hungry Data Centers Stretch the Grid

Silicon Valley’s appetite for computing power could soon outstrip the power grid’s capacity. Deloitte projects “AI data center power demand will surge from 4 gigawatts in 2024 to 123 gigawatts by 2035,” a more than thirty‑fold jump. BloombergNEF warns that the load could rise to “7‑20 percent of total U.S. electricity demand by 2030,” up from 3.5 percent today.

The largest new campuses will “require up to 2 gigawatts of power — the output of a major nuclear plant,” and some clusters could hit 5 GW. Meanwhile, “over 2.6 terawatts of generation await grid connection, with waits that now average five years.” The Action Plan acknowledges these bottlenecks but offers few concrete solutions beyond streamlining the siting process.

A Talent Gap — and Bidding War — Widens

Tech giants can buy GPUs, but not engineers. The world faces “about 4.2 million unfilled AI positions, with only 320,000 qualified developers.” IBM puts the 2024 hiring gap at “50 percent, with 87 percent of firms struggling to hire and time‑to‑fill averaging 142 days.”

By 2027, “up to 700,000 U.S. workers may need reskilling”. Pay follows scarcity: companies now “offer up to four times entry‑level salaries,” with AI compensation rising 11 percent a year since 2019.

Federal Buying Rules Rewritten

In April, the Office of Management and Budget issued new guidance that eliminates previous restrictions on AI bias and discrimination concerns.” Agencies must now “maximize the use of American‑made AI while avoiding vendor lock‑in through data portability.” Removed biometric‑system mandates and incident‑reporting rules should speed procurement but raise civil liberties alarms. Officials also instruct departments to create “high‑impact AI” categories with lighter oversight, replacing dual safety‑and‑rights tiers. 

Key Takeaways:

• Trump’s 90‑point AI plan sweeps away Biden‑era safety rules and enshrines U.S. dominance as policy.
• Three executive orders ban “woke AI,” boost exports, and root out deregulation in law.
• AI chip export controls tightened, then loosened, underscoring policy volatility.
• Data‑center power demand could soar 30‑fold by 2035, straining the U.S. grid.
• A severe talent shortage is inflating salaries and slowing deployment worldwide.