OpenAI urges expansion of CHIPS Act tax credit to cover AI infrastructure
OpenAI has asked the administration to expand the CHIPS Act's Advanced Manufacturing Investment Credit (AMIC) beyond semiconductor fabs to include AI servers, data centers, and critical grid hardware. The filing was submitted to the Office of Science and Technology Policy's long-term AI policy consultation and posted publicly by the company.
The request is framed as a way to cut capital costs, speed buildouts, and shore up strained supply chains. It also separates tax credits from loan guarantees: OpenAI says it is not asking for direct subsidies or a government "backstop" for its own operations, and that any explored guarantees were tied to semiconductor partners.
What AMIC covers today
AMIC is a 25% investment tax credit under Section 48D, finalized by Treasury in 2024. It currently applies to semiconductor manufacturing and tooling. Expanding it to AI data centers and servers would likely require Congress to amend the statute, even though Treasury has some interpretive discretion.
IRS: Advanced Manufacturing Investment Credit (Section 48D)
White House Office of Science and Technology Policy
What OpenAI is asking to include
- Domestic production of AI servers and related systems.
- Construction of AI-focused data centers.
- Grid equipment with long lead times: high-voltage transformers, HVDC converters, and transmission components.
- Additional grants, loans, and guarantees to scale manufacturing of critical components.
Why this matters for government
The industry is racing to secure power, hardware, and sites for large inference clusters. OpenAI's leadership has floated U.S. demand that could require up to 100 GW of new generation, with a single next-gen project ("Stargate") reportedly planning for as much as 5 GW.
That scale touches multiple public responsibilities: grid reliability, permitting, domestic manufacturing, workforce development, environmental review, and fair allocation of limited transmission capacity. It also intersects with U.S. competitiveness as China and Gulf states build aggressively.
Key policy constraints
- Statutory scope: AMIC is defined by Congress; broadening to data centers and AI systems likely needs legislation.
- Program overlap: Interactions with other credits and grants (e.g., manufacturing and energy programs) require coordination to avoid double-dipping.
- Guardrails: Prevailing wage, apprenticeship, domestic content, and cybersecurity standards may be necessary for parity with existing federal incentives.
- Budget impacts: CBO scoring, subsidy stacking, and timing of outlays versus tax receipts will be central to any proposal.
Supply chain and grid realities
- Lead times for large power transformers and HVDC equipment often exceed 24-36 months.
- Transmission planning and interconnection queues are congested, creating delays that dwarf server deployment cycles.
- Data center siting increasingly hinges on access to firm power, water, and high-capacity fiber, not just land and tax abatements.
Questions decision-makers may want answered before expanding AMIC
- Scope: How is an "AI server" defined for eligibility, and how do we prevent rebranding of commodity servers to qualify?
- Additionality: What evidence shows the credit changes build timing, siting, or domestic production-versus subsidizing projects that would happen anyway?
- Domestic content: What thresholds meaningfully build U.S. capacity without stalling deployments?
- Grid integration: Should eligibility hinge on co-investment in transmission, on-site firming, or demand flexibility to reduce system stress?
- Efficiency and sustainability: Will minimum PUE, water-use, and waste-heat standards be required?
- Regional balance: How do credits support buildout beyond a few metros while aligning with reliability needs and community benefits?
- Security: What cybersecurity and supply-chain integrity controls are necessary for federally supported AI infrastructure?
Agency considerations if Congress moves
- Define measurable criteria for AI data centers and server manufacturing eligibility to keep the credit targeted.
- Coordinate with DOE, FERC, and state regulators on transformer capacity, HVDC corridors, and interconnection process reforms.
- Set clear reporting for grid impacts, job creation, and domestic content to enable oversight and course correction.
- Align with environmental review timelines and community engagement requirements to de-risk siting.
- Prepare procurement templates that incorporate labor, security, and efficiency standards without adding unnecessary delay.
Optics and precedent
The filing distances itself from direct subsidies to OpenAI and focuses on sector-wide infrastructure. Even so, it will raise questions about preferential support for energy-intensive private workloads and the fairness of subsidizing data centers alongside fabs.
If Congress decides to act, clarity on definitions, guardrails, and measurable public benefits will determine whether an expansion of AMIC is seen as a strategic investment or a misaligned subsidy.
Workforce and capability building
Public teams charged with evaluating AI infrastructure policy may benefit from structured upskilling on AI systems, data center fundamentals, and procurement. For role-based learning paths, see Complete AI Training: Courses by Job.
Your membership also unlocks:
