Trump's $874 Million CHIPS Act Deal: 7 Companies, Government Equity, and What's Actually Investable

The Deal at a Glance

Commerce Secretary Howard Lutnick's department announced non-binding letters of intent — not final, executed contracts — to provide up to $874 million in CHIPS and Science Act incentives across seven companies. All seven work somewhere in the AI compute supply chain: the optics, memory, packaging, and exotic computing hardware that determine how fast and how cheaply AI chips can be built and connected together.

The structural twist is the same one Washington used with Intel earlier in 2026: rather than a no-strings-attached grant, the government is converting part of each award into a minority, non-controlling equity stake in the recipient company. Lutnick framed the goal as strengthening "domestic supply chains and accelerating AI technologies" — but the mechanism itself, government-as-shareholder, is now the more interesting story for investors than the underlying technology grants.

The Seven Recipients and Their Awards

The awards range from $30 million to $245 million and sum to the full $874 million headline figure:

Company-by-Company Breakdown

This is the single most important fact for retail investors trying to trade this news: five of the seven recipients — Kepler, Multibeam Corporation, Extropic, Thintronics, and OBSIDIA Semiconductors — are privately held, venture-backed startups. There is no ticker symbol, no public filing, and no way for a retail investor to buy shares. Any headline suggesting a broad "basket" of tradable winners from this announcement is overstating what's actually accessible.

• GlobalFoundries (GFS) is the largest award recipient at up to $300M and is already a familiar CHIPS Act name — it also holds a $1.5B grant and operates a strategically important mature-node fab in Malta, NY that supplies defense and automotive customers
• Aeluma (ALMU) is a Nasdaq-listed micro-cap that received the smallest award, up to $30M — its photonic-chip-on-silicon technology is unproven at commercial scale, and the stock should be treated as a speculative, highly volatile small-cap rather than a core position
• The five private recipients may eventually IPO or get acquired, at which point the government's equity stake — and any future public listing — would become directly relevant to public markets, but that is a multi-year, uncertain path
• The more durable way to invest in this theme is through already-public companies operating in the same technology categories at commercial scale, covered below

This $874 million deal did not happen in isolation. It's the latest addition to a rapidly growing government corporate equity portfolio that now spans roughly 30 companies and an estimated $27 billion in deployed capital, held across at least four federal agencies: Commerce (around 17 deals), the Department of Defense (around 7), the Development Finance Corporation (around 6), and the Department of Energy (around 2).

The Intel position remains the benchmark case for how this model can play out: what started as a converted CHIPS Act grant into roughly a 10% equity stake now carries an unrealized paper gain reported north of $70 billion as INTC shares have rallied through 2026. That single outcome is doing a lot of work to make the equity-for-grant model look politically and financially attractive going forward — and likely explains why the July 2026 deal used the same structure rather than a straight grant.

This also followed a smaller precedent set in May 2026, when the government distributed nearly $2 billion across nine companies including a $1 billion award to IBM tied to quantum computing manufacturing — meaning the July deal is the second major AI/advanced-computing-focused equity-for-grant round in three months, not an isolated event.

Since five of the seven direct recipients are off-limits to public investors, the cleanest way to position around this announcement is through established public companies that already compete commercially in the same technology categories the CHIPS award is trying to seed.

The seven awards are not random — they cluster tightly around the specific engineering bottlenecks slowing down AI compute scaling today:

Co-Packaged Optics ($300M to GlobalFoundries, plus Aeluma's $30M photonics award)

As AI clusters scale to tens of thousands of GPUs, the copper wiring connecting chips becomes a power and bandwidth bottleneck. Co-packaged optics fuses optical transceivers directly onto the chip package, cutting power consumption per bit moved. This is widely viewed as a necessary technology shift for the next generation of AI data centers — both GlobalFoundries and Aeluma are placing bets on different manufacturing approaches to the same problem.

AI Memory ($245M to Kepler)

Modern AI accelerators are frequently memory-bandwidth-constrained rather than compute-constrained — a GPU can often process data faster than memory can feed it. Kepler's 3D/ferroelectric approach is a moonshot alternative to the HBM (high-bandwidth memory) architecture that Micron, SK Hynix, and Samsung already ship commercially.

Advanced Packaging ($140M to Multibeam)

Stacking multiple chiplets into a single package — the same broad category as TSMC's CoWoS packaging used for NVIDIA's most advanced GPUs — has become a capacity bottleneck across the entire AI hardware industry. Multibeam's electron-beam approach is an attempt at a lower-cost, higher-throughput alternative to existing packaging methods.

Exotic Computing & Supply Chain Security ($75M to Extropic, $50M to Thintronics, $34M to OBSIDIA)

Extropic's thermodynamic computing bet is the most speculative award on the list — a fundamentally different computing paradigm rather than an incremental improvement. Thintronics addresses a materials-science input (dielectrics) that feeds into both the optics and packaging categories above. OBSIDIA's counterfeit-detection technology addresses a supply-chain-security concern that has grown alongside the broader push to reshore and secure chip production.

• The Intel precedent shows the equity-for-grant model can generate enormous taxpayer returns if the underlying company performs — a $70B+ unrealized gain on one stake changes the political calculus for repeating the structure
• Targeting specific bottleneck technologies (optics, memory, packaging) rather than broad manufacturing capacity suggests a more technically sophisticated allocation of CHIPS Act dollars in this later phase of the program
• GlobalFoundries gets a second funding stream on top of its existing $1.5B CHIPS grant, reinforcing its position as a domestic foundry the government is committed to supporting long-term
• For public-market investors, the categories being funded (co-packaged optics, AI memory, advanced packaging) are the same bottlenecks that Broadcom, Micron, Coherent, and Amkor are already monetizing at commercial scale — the government spending is a tailwind confirming the thesis, not a prerequisite for it
• A rapidly growing, increasingly familiar policy playbook (30 companies now) suggests this is durable industrial policy rather than a one-off, reducing uncertainty about whether similar deals continue

• Five of seven recipients are private — there is effectively no direct, liquid public-market way to play most of this announcement, meaning traders bidding up tangential small caps on the headline risk chasing an illusion of exposure
• Government-as-shareholder in dozens of companies raises real concerns about market distortion and winner-picking — critics warn the approach could crowd out private capital or create moral hazard where companies court government funding over commercial validation
• Opacity is a genuine problem: private-company equity stakes carry none of the disclosure requirements of public holdings, making it difficult for taxpayers or investors to track the value, terms, or performance of most of the roughly 30-company portfolio
• Letters of intent are non-binding and pending further review — some or all of these seven deals could be renegotiated, downsized, or fall through entirely before final agreements are signed
• Aeluma (ALMU) is a micro-cap with unproven commercial-scale manufacturing — the $30M award validates the technology direction but does not de-risk execution, and the stock can be extremely volatile on news flow alone
• A change in administration or a shift in political winds could unwind or freeze the broader equity-stake program, introducing policy-continuity risk for any company whose valuation has come to depend partly on government backing

Bottom Line

This $874 million deal is less important as a stock-picking event than as confirmation of a structural shift in US industrial policy: the government is no longer just writing grant checks for chip manufacturing — it is becoming a minority shareholder across dozens of companies spanning semiconductors, critical minerals, energy, and advanced manufacturing, with total deployed capital approaching $27 billion.

For most investors, the actionable takeaway is narrow but clear: GlobalFoundries (GFS) is the one large, liquid, direct beneficiary, and Aeluma (ALMU) is a small, speculative, direct beneficiary. Everyone else looking to invest in the underlying themes — AI memory bottlenecks, co-packaged optics, advanced packaging — is better served buying the established commercial leaders already executing at scale: Micron, Broadcom, Coherent, and Amkor. The Intel precedent suggests this policy model has real financial upside if it keeps working, but it also means investors should watch future letters of intent from Commerce, Defense, the DFC, and Energy as a recurring, market-relevant event rather than a one-time headline.

Explore the Public Companies Involved

Dig into our analysis tools for the tradable names from this story and the broader theme: