IPO – Stacking Trades

Cerebras Systems Wants to Test the AI Chip Market Before Nvidia Does It for Them

Stacking Trades — Mon, 06 Apr 2026 18:49:39 +0000

The last time Cerebras Systems tried to go public, it withdrew its registration statement in October 2025 — days after closing a funding round — citing an unresolved national security review of a minority investment from Abu Dhabi-based technology firm G42. The optics were not ideal. The company’s first prospectus had revealed that a single foreign customer represented roughly 87% of its revenue through the first half of 2024, and federal regulators wanted to understand what that relationship meant for sensitive American compute infrastructure.

That chapter is closed. G42 has since been removed from Cerebras’s primary shareholder structure to satisfy U.S. regulators, and the company has spent the months since building a customer base that looks nothing like the one in that first S-1. In January 2026, Cerebras signed a $10 billion compute deal with OpenAI, pledging 750 megawatts of computing capacity through 2028. In March, it announced a partnership with Amazon Web Services to deploy its CS-3 systems inside AWS data centers, available through Amazon Bedrock. The company is now valued at $23.1 billion after a February Series H round and is targeting a roughly $2 billion raise on the Nasdaq, with Morgan Stanley as lead underwriter, according to Bloomberg.

The public S-1 has not yet been filed as of this writing. But the architecture of the deal — the timing, the customer lineup, the deliberate sequencing of announcements — reads like a company that understands exactly what a prospectus needs to say.

The Chip That Doesn’t Fit the Nvidia Model

To understand why Cerebras matters to investors, you need to understand why it is structurally different from every other AI chip company trying to go public right now. Nvidia’s dominant GPU architecture works by connecting hundreds or thousands of discrete chips — each physically small — through high-bandwidth memory and fast interconnect. The bottleneck in that approach is data movement: getting information from one chip to another, from memory to processor, fast enough to keep pace with the model’s demands.

Cerebras built the WSE-3 from the other direction. The chip is a single processor the size of an entire 300mm silicon wafer — roughly 56 times the physical area of Nvidia’s H100. It contains 4 trillion transistors, 900,000 AI-optimized cores, and 44 gigabytes of on-chip SRAM with 27 petabytes per second of internal memory bandwidth. Because the model weights live on the chip itself rather than in external memory, there is no bottleneck to solve. The machine simply runs faster — particularly in inference tasks where an AI system is generating responses to live queries, rather than training on new data.

The practical result: Cerebras delivered Llama 4 Maverick inference at more than 2,500 tokens per second per user on its CS-3 system, compared to roughly half that on Nvidia’s flagship DGX B200 Blackwell running the same 400-billion parameter model. For applications like agentic coding tools — where a developer is waiting for multi-step AI reasoning in real time — that difference is meaningful.

“Every customer large or small is on AWS, from individual developers to the largest banks in the world. The deal will make it easy as a click to get on Cerebras.”
— Andrew Feldman, CEO, Cerebras Systems, Reuters, March 13, 2026

The Amazon Deal Changes the Distribution Equation

For most chip startups, hardware reach is the hardest problem. You can build the fastest processor in the world and still lose if your customers can’t access it through the infrastructure they already use. The AWS partnership, announced March 13, addresses that directly. Under the arrangement, Cerebras CS-3 systems sit inside AWS data centers and operate alongside Amazon’s own Trainium3 chips in a so-called disaggregated inference architecture — Trainium handles the prefill stage of a query, Cerebras handles the decode. AWS calls the result five times the high-speed token capacity in the same hardware footprint. The service, running on Amazon Bedrock, is expected to launch in the second half of 2026.

The significance for Cerebras is distribution at a scale no startup can build independently. AWS serves customers ranging from individual developers to global financial institutions. When David Brown, Vice President of Compute and ML Services at AWS, said publicly that the Trainium-Cerebras solution will deliver “inference that’s an order of magnitude faster and higher performance than what’s available today,” that is not a press release formality. It is a co-endorsement from the world’s largest cloud provider, delivered weeks before an IPO roadshow.

Cerebras has also inked IBM and the U.S. Department of Energy as customers, alongside OpenAI, Cognition, and Mistral. The customer concentration risk that sank the first S-1 story has been structurally dismantled. The question is whether the new customer roster can support the valuation.

The Valuation Math Is Tight

At the $23 billion figure established in the February Series H, Cerebras would debut as one of the ten largest semiconductor IPOs in history, priced ahead of its current revenue. Estimated 2025 revenues exceeded $1 billion according to multiple analyst reports, but the company’s cost structure — proprietary water-cooled hardware, TSMC wafer manufacturing, and a software stack that requires developers to leave Nvidia’s CUDA ecosystem — is not cheap to operate.

The CUDA problem is worth understanding. Nvidia’s developer ecosystem is the deepest competitive moat in the chip industry. Tens of thousands of enterprise AI teams write code specifically for CUDA; switching to Cerebras’s software stack requires retraining and re-tooling. The company’s inference API — which lets developers access wafer-scale performance through a standard cloud interface without buying hardware — is designed to lower that barrier. But it does not eliminate it. For institutional investors pricing the IPO, the question is how many enterprise customers will opt for Cerebras performance at a premium over Nvidia compatibility at a discount.

The Amazon integration changes that calculus somewhat. If developers can access Cerebras hardware through a standard Bedrock API call — the same interface they already use for other AWS AI services — the switching cost drops considerably. That may be the single most important structural fact about the March 13 announcement, and it is likely to feature prominently in the S-1.

What the Second Attempt Gets Right

Cerebras has learned from the timing mistake of the first filing. The original S-1 landed in September 2024 into a national security review it could not resolve quickly. The company tried to wait it out, raised capital to extend its runway, and ultimately withdrew. This time, the regulatory pathway was cleared before the filing, the key customer relationships were announced in sequence — OpenAI in January, Amazon in March — and the underwriter was selected before the formal S-1 submission.

The IPO window for Q2 2026 is not guaranteed to stay open. Market volatility and macro uncertainty can compress or shut the calendar quickly, as the broader IPO market has demonstrated multiple times in the last 18 months. The company’s Nasdaq ticker reservation — CBRS — has been held since the first filing. Whether it gets used in April or slides to June will depend on when the public S-1 drops and how investor appetite looks after the bank earnings season that begins the week of April 13.

What is clear is that Cerebras is no longer asking the market to fund a technical bet on an unproven architecture. It is asking the market to value a company that OpenAI, Amazon, IBM, and the U.S. Department of Energy have already paid to use.

WHAT TO WATCH NEXT

The public S-1 filing on SEC EDGAR — expected in late April or early May before any roadshow; it will contain the first audited revenue figures, cost structure, and TSMC manufacturing dependency disclosure.
AWS Bedrock launch date for the Trainium-Cerebras disaggregated service — the second-half 2026 window is wide; an earlier-than-expected rollout would strengthen the IPO narrative heading into pricing.
Nvidia’s response — Reuters reported Nvidia is expected to combine its own GPU chips with Groq (acquired for $17 billion in December 2025) in a similar disaggregated inference architecture. That announcement, if it arrives before Cerebras prices, directly affects how investors frame the competitive risk.
OpenAI contract execution milestones — the $10 billion agreement runs through 2028, but delivery is staged; any disclosure of compute capacity actually deployed versus committed will be the most meaningful revenue signal in the S-1.
TSMC wafer allocation — Cerebras uses nearly an entire 300mm wafer per chip and competes directly with Apple and Nvidia for TSMC manufacturing capacity; any tightening in that supply chain is a direct production risk.

SpaceX’s Confidential Filing Is the Starting Gun, Not the Finish Line

Stacking Trades — Thu, 02 Apr 2026 22:34:11 +0000

On April 1, 2026, SpaceX submitted its confidential draft registration to the U.S. Securities and Exchange Commission, setting in motion what could be the largest initial public offering in stock market history. Bloomberg reported the filing first. CNBC, Reuters, and the Wall Street Journal confirmed it within hours. The company is targeting a June listing. The valuation cited by multiple outlets is $1.75 trillion, with a capital raise that could reach $75 billion — more than double the $29 billion Saudi Aramco raised in 2019, which currently holds the record.

None of that is the story. The story is what investors don’t have yet.

A confidential filing — standard practice for large listings under the JOBS Act — allows SpaceX to work through SEC comments privately before its financials become public. Under SEC rules, the company must publicly file its full prospectus at least 15 days before the investor roadshow begins. With a June target, that prospectus is expected to land in April or May. Until it does, every valuation figure in circulation is a number derived from secondary market trades, analyst estimates, and sources who spoke to reporters anonymously. No audited revenue breakdown. No subscriber acquisition cost. No disclosed cash burn attributable to xAI. No formal cap table post-merger.

The Starlink Question Is the Only One That Matters

The $1.75 trillion valuation is not primarily a bet on rockets. Starlink ended 2025 with 9.2 million subscribers and generated over $10 billion in revenue, with analysts projecting that figure could reach anywhere from $15.9 billion to $24 billion in 2026 depending on subscriber trajectory and pricing mix. The wide range in analyst projections — a gap of more than $8 billion in a single year — is itself the problem. Nobody has seen the actual unit economics.

What does an average Starlink subscriber generate per month, net of satellite operations, customer support, and terminal subsidies? How much of revenue is consumer versus maritime versus aviation versus government? What is churn? These are the figures that underpin any credible discounted cash flow model, and right now none of them exist in public form. SpaceX’s total 2025 revenue has been estimated at $15 billion to $16 billion with roughly $8 billion in profit, but those are figures sourced from Reuters citing unnamed sources, not from audited financials.

For a company targeting a valuation higher than every S&P 500 constituent except Nvidia, Apple, Alphabet, Microsoft, and Amazon, the margin of uncertainty here is unusual.

The xAI Integration Adds an Entirely New Layer of Risk

In early February 2026, SpaceX completed an all-stock acquisition of xAI, Musk’s artificial intelligence startup, in a deal that valued SpaceX at $1 trillion and xAI at $250 billion — the largest merger in history by combined entity size. The stated rationale was building orbital data centers: using Starlink’s satellite constellation to move AI compute workloads into space, where solar power is abundant and cooling is free.

The concept is technically ambitious. Whether it is financially credible is something the prospectus will need to answer. At the time of the merger, xAI was burning roughly $1 billion per month, according to Bloomberg, as it raced to build out infrastructure against OpenAI and Google. That burn rate, absorbed by SpaceX at the moment of the filing, now sits inside the entity investors are being asked to value at $1.75 trillion.

“The valuation jump from $1.25 trillion at merger close to $1.75 trillion at filing reflects expectations around the combined entity’s space and AI ambitions — not disclosed financials.”
—TheStreet, April 2, 2026

Then there is the talent picture. xAI co-founders Jimmy Ba and Tony Wu departed in February, shortly after the merger closed. By early March, Zihang Dai and Guodong Zhang had also left. Of the 12 people who co-founded xAI with Musk in 2023, only two remain — Manuel Kroiss and Ross Nordeen. Musk publicly acknowledged on X that xAI “was not built right first time around” and needed to be rebuilt from its foundations. That statement, made weeks after the largest merger in history closed at a $250 billion xAI valuation, will almost certainly surface in the prospectus’s risk factors — and will need to be reconciled with the financial picture investors are presented.

The Governance Structure Has No Precedent

SpaceX’s governance heading into a public offering is unlike any company that has come before it. Musk simultaneously holds the CEO role at SpaceX, Tesla, and leads DOGE in a senior advisory capacity. Tesla disclosed that it invested $2 billion of shareholder money into xAI in its Q4 2025 earnings report — and then SpaceX acquired xAI days later. Legal analysts tracking the deal note that consummating a conflicted merger between founder-controlled private companies before going public defers — but does not eliminate — the scrutiny around related-party dynamics. That scrutiny will arrive when the S-1 is public and securities lawyers start reading it.

SpaceX is also reportedly considering a dual-class share structure that would preserve insider voting control, and plans to allocate up to 30% of shares to retail investors — roughly three times the typical allocation in a large IPO. The retail allocation is an unusual move that deserves attention: it broadens the investor base and generates enthusiasm, but it also introduces a class of shareholders with limited governance recourse under a dual-class structure.

What the 21-Bank Syndicate Actually Signals

SpaceX has lined up at least 21 banks for this offering, with Goldman Sachs, JPMorgan Chase, Morgan Stanley, Bank of America, and Citigroup in senior underwriting roles. The deal is internally codenamed “Project Apex.” A syndicate of this size indicates that demand management is already the primary concern — no single bank wants the concentration risk of placing $75 billion in a single offering. It also means the roadshow, when it happens, will be one of the most heavily orchestrated investor events in market history.

One detail worth noting for index-aware investors: Nasdaq recently updated listing rules in a way that could allow SpaceX to join the Nasdaq 100 within 15 days of listing. That would trigger forced buying from index-tracking funds — a mechanical demand wave that has nothing to do with the underlying fundamental case. Investors who intend to evaluate SpaceX on its merits should be aware that initial price behavior after listing may reflect index inclusion mechanics more than actual price discovery.

Stacking Trades has covered the broader conditions shaping the IPO market this spring, including how macro headwinds in early 2026 rattled the IPO window — context that matters when assessing whether a $1.75 trillion listing can clear the market in June regardless of how good the underlying story is.

The Number That Investors Actually Need

The prospectus, when it arrives, will contain information that no source has yet provided: audited revenue by segment, Starlink subscriber economics, the precise mechanics of the xAI acquisition and how xAI’s assets and liabilities are carried on SpaceX’s balance sheet, the terms of the dual-class structure, and the risk factors Musk’s lawyers have deemed material enough to disclose to the investing public.

That document will be the first moment at which a credible, independently verified number can be placed next to the $1.75 trillion figure. Until then, every model being built against that valuation is working from incomplete information. That does not make the offering unattractive. It means the starting gun has been fired — and the race to price it properly has not yet begun.

WHAT TO WATCH NEXT

The public S-1 prospectus: Expected in April or early May. The Starlink revenue breakdown, per-subscriber economics, and xAI integration terms are the first numbers worth modeling against the $1.75 trillion target.
xAI’s disclosed cash burn post-merger: At roughly $1 billion per month at acquisition close, any updated figure in the prospectus will materially affect the combined entity’s free cash flow profile.
Dual-class share structure details: The voting power Musk retains — and what that means for board independence and related-party transaction governance — will be the section securities lawyers read first.
Nasdaq 100 inclusion timing: If confirmed at listing, forced buying from index funds could distort early price discovery. Watch for Nasdaq’s formal ruling.
OpenAI and Anthropic IPO timelines: Bloomberg has reported both companies are weighing public offerings in 2026. SpaceX’s pricing and reception will set the reference point for that entire class of mega-listings.

Private Credit’s Retail Reckoning Is a BDC Problem — Not a Crowdfunding One

Stacking Trades — Thu, 02 Apr 2026 16:39:17 +0000

The headlines out of the private credit industry this quarter have been stark. Blackstone allowed investors to pull a record 7.9% of assets from its flagship $82 billion private credit fund — well above the standard 5% quarterly cap — after fielding roughly $3.8 billion in withdrawal requests. Blue Owl had it worse. Its technology-focused business development company saw redemption requests hit 15% of net asset value; a separate fund was gated entirely, with Blue Owl replacing quarterly tender offers with rateable return-of-capital distributions.

The industry response has been swift and, at times, defensive. Blue Owl argued it was “accelerating the return of capital,” not restricting it. Blackstone President Jon Gray said market “noise” had driven nervous retail investors toward the exits, pointing to BCRED’s 9.8% annualized return since inception as evidence the underlying portfolio remains sound. CALPERS, the California pension giant, was among the institutional buyers that swooped in to acquire loan assets Blue Owl sold at book value to fund the redemptions.

For investors who follow private market deal flow — including those active in Regulation A+ and Regulation Crowdfunding offerings — this story carries a real risk of misinterpretation. The BDC liquidity crisis is structurally specific. It does not apply to the exemption-based crowdfunding market, and conflating the two could lead to decisions that don’t match the actual exposure.

What Is Actually Breaking Down — and Why

Business development companies are a federally chartered structure, created by Congress in 1980, that package private credit — primarily loans to middle-market companies — and sell shares to retail investors. The key feature that makes them attractive also makes them fragile in stress conditions: they offer periodic liquidity. Under a typical non-traded BDC structure, investors can request to redeem up to 5% of net asset value per quarter. That cap exists precisely because the underlying loans are illiquid. Managers need runway to sell assets, collect repayments, or arrange credit lines before cash can move out the door.

When retail investors — less patient than institutional allocators who understand and accept lockups — become anxious and request more than 5% at once, the fund faces a mismatch between what it has promised and what it can deliver without forcing asset sales. That is what happened here. Blackstone resolved it by injecting roughly $400 million in firm and employee capital to offset excess redemptions, maintaining its record of meeting 100% of requests. Blue Owl’s OBDC II fund lacked that solution set and moved instead to gate the structure entirely.

“The product expanded faster than the communication infrastructure that should have accompanied it. That is the root cause of the current confidence issue.”
— Private Markets Insights, March 2026, citing Bloomberg Invest 2026 panel discussion

Morningstar, in its analysis of the BCRED situation, noted that Blackstone’s ability to use its own balance sheet as a liquidity buffer is not available to all managers. Smaller private credit funds offering semi-liquid structures but lacking a $82 billion sponsor’s resources are materially more exposed to the same dynamic. The lesson is not that private credit is broken. BCRED has outperformed leveraged loans by 360 basis points since inception. The lesson is that the structure carrying it to retail investors has stress points that were not clearly communicated.

Why Reg A+ and Reg CF Work Differently

Regulation A+ and Regulation Crowdfunding offerings are not semi-liquid funds. They are exempt securities offerings governed by the SEC under the Securities Act of 1933 — specifically, Title III of the JOBS Act for Reg CF and Regulation A for what is commonly called Reg A+. The investor relationship in these structures is categorically different from a BDC investor’s relationship with a redemption window.

When an investor puts capital into a Reg CF offering, they are buying an equity stake — or, in some cases, a revenue-share or debt instrument — in a private company. There is no fund manager offering periodic redemptions. There is no quarterly tender offer. There is no liquidity window that can be gated. The investor holds a security in a company, and that security becomes liquid only through a defined event: a company IPO, an acquisition, a secondary market transaction, or a structured repurchase. Reg A+ securities, notably, may be sold immediately to any investor without the one-year holding period that applies to Reg CF securities. That makes them more liquid by design, not less.

The table below summarizes the structural differences that matter for investors trying to separate the BDC story from the broader private markets story.

Structure

Liquidity promise

Redemption risk

Investor relationship

Non-traded BDC
(e.g., BCRED, OTIC)

Quarterly tender, typically 5% of NAV

High — mismatch between promise and underlying illiquidity

Fund shareholder with periodic redemption rightsInvestor relationship

Reg CF offering

None — exit via liquidity event only

None — no redemption window to gate

Direct equity/debt holder in issuer

Reg A+ (Tier 2) offering

Immediate secondary market sale allowed

None — securities are freely transferable

Direct securities holder; may trade without restriction

Interval fund / evergreen BDC

Periodic (quarterly or semi-annual) limited redemption

Medium — gating possible under stress; structure-dependent

Fund investor with contractual liquidity window

The DOL Rule That Changes the Stakes

All of this arrives in the same week the Department of Labor published a proposed rule that would make it materially easier for 401(k) plans to include private market investments. The rule, released March 30, creates a process-based safe harbor for fiduciaries who want to add alternatives — including private equity, private credit, and cryptocurrency — to their defined-contribution menus. Labor Secretary Lori Chavez-DeRemer said the rule would “show how plans can consider products that better reflect the investment landscape as it exists today.”

With roughly $12 trillion in defined-contribution assets sitting largely outside private markets, the opportunity is significant. So is the timing problem. The DOL is proposing to pipe more retail capital into structures that have just spent two months generating headlines about redemption failures and communication breakdowns. Critics were immediate. Sen. Elizabeth Warren described it as a mechanism for “Wall Street buddies” to access retirement savings. The Employee Benefit Research Institute flagged that current market instability complicates the outlook for adoption regardless of what the rule says.

More substantively, legal analysts at TD Cowen said they remain skeptical the rule will move fiduciaries off the sidelines without court confirmation that its safe harbor language holds against litigation. That is not a fast-moving variable. Employers that were sued over their 401(k) menus do not move quickly in response to proposed rules. They move in response to finalized rules — and then only after attorneys confirm the shield actually works.

What Issuers Should Take From This

For companies running Reg A+ or Reg CF campaigns, the noise around private credit redemptions is not your risk — but it is your opportunity. The BDC episode is teaching a generation of new private market investors a lesson about structure and liquidity. Some of those investors will be more cautious going forward. Others will be looking for private market exposure that does not carry a redemption-window risk, which is precisely what equity crowdfunding offers at the design level.

The sophistication gap that Blue Owl’s co-founder acknowledged — the gap between the pace of product expansion and the pace of investor education — is not unique to BDCs. Reg CF issuers that communicate clearly about what their securities are, when liquidity can be expected, and what milestones govern that timeline are positioned to attract investors who just received an expensive tutorial on what happens when those things go unexplained. The issuers that treat their post-raise investor communications with the same discipline they apply to the campaign itself are the ones who will benefit from the BDC sector’s credibility problem.

WHAT TO WATCH NEXT

Whether any additional non-traded BDC gates redemptions in Q2 2026, particularly from managers without Blackstone-scale balance sheets to absorb excess requests.
The DOL’s public comment period outcome on its 401(k) alternatives rule — the 60-day window will determine how quickly fiduciaries can act, and whether the rule survives legal challenge.
Blue Owl’s orderly liquidation of OBDC II — if the fund returns capital at book value to all shareholders by mid-year, it becomes a model case study; if asset sales produce discounts, the narrative shifts toward credit quality concerns.
SEC activity on Reg CF cap reform — the pending petition to raise the Reg CF limit from $5 million to $20 million gains relevance if institutional interest in the crowdfunding channel increases as BDC credibility falters.
Whether any Reg A+ or Reg CF platforms explicitly position against BDC risk in their investor communications — that marketing angle has not been used publicly yet and may become a differentiator.

The Nuclear IPO That AI Built: Inside X-Energy’s Bid to Go Public

Stacking Trades — Thu, 02 Apr 2026 02:05:30 +0000

X-energy, Inc. filed its S-1 registration statement with the SEC on March 20, 2026, setting up what could be the first pure-play advanced nuclear company to reach the public markets in a generation. The company intends to list on the Nasdaq Global Select Market under the ticker “XE,” with J.P. Morgan, Morgan Stanley, Jefferies, and Moelis & Company acting as lead bookrunners. Renaissance Capital estimates the raise at approximately $300 million, though the number of shares and price range have not yet been determined.

The timing is not coincidental. Every major technology company with a balance sheet is currently trying to solve the same problem: where do you get gigawatts of firm, always-on, carbon-free power for AI data centers that run every hour of every day? Renewables can’t do it alone. Natural gas doesn’t satisfy the zero-carbon commitments most of these companies have made. That leaves nuclear — specifically, a new generation of smaller, faster-to-build reactors that don’t require the decade-long permitting fights of their 1970s predecessors. X-energy is making a direct bid to be that answer.

What X-Energy Actually Builds

Founded in 2009 by aerospace entrepreneur Kam Ghaffarian and now led by J. Clay Sell — a former U.S. Deputy Secretary of Energy — X-energy designs the Xe-100, a high-temperature gas-cooled small modular reactor that produces 80 megawatts of electricity per unit, scalable to a four-pack generating 320 megawatts. The reactor runs on TRISO-X fuel, a proprietary tri-structural isotropic particle fuel the Department of Energy has called “the most robust nuclear fuel on Earth.” Unlike conventional light-water reactors, TRISO fuel cannot melt down in the traditional sense — the ceramic coating on each fuel particle provides its own containment.

The reactor design’s key commercial appeal is modularity. X-energy describes its Xe-100 as road-shippable, with accelerated construction timelines and more predictable costs than conventional nuclear. Whether that promise holds when the first reactor actually goes into the ground at Dow Inc.’s Seadrift, Texas site remains to be seen. But it is the premise on which the company has assembled a remarkable roster of partners.

Amazon, Dow, and the Orderbook Behind the Filing

The commercial logic of X-energy’s IPO rests heavily on two anchor relationships. The first is Dow Inc., which signed a joint development agreement in 2023 to deploy a four-reactor Xe-100 plant at its Seadrift Gulf Coast chemical complex — the first grid-scale advanced nuclear reactor deployed to serve an industrial site in North America. The NRC docketed the construction permit application in May 2025 and published an 18-month review timeline, with environmental review running concurrently. Construction is expected to begin in 2026.

The second is Amazon, which anchored a $500 million Series C-1 round in October 2024 through its Climate Pledge Fund. Beyond the equity investment, Amazon has options to deploy more than 5 gigawatts of Xe-100 projects across the United States by 2039, beginning with the Cascade Advanced Energy Facility in Washington state, built in partnership with utility Energy Northwest adjacent to the Columbia Generating Station. A separate agreement with UK energy firm Centrica outlines plans for roughly 6 gigawatts of Xe-100 capacity in Britain.

X-energy’s total disclosed orderbook now exceeds 11 gigawatts across approximately 144 advanced reactors, according to the company’s own disclosures at the time of its Series D close. That is a pipeline, not a backlog — no revenue flows until reactors are built and operating. But for a pre-commercial nuclear company, having Amazon and Dow as co-development partners is a categorically different risk profile than having nothing but a design.

“X-energy’s technology will help satisfy historically unprecedented electricity demand growth, driven by the development of AI and associated data center infrastructure.”
—X-energy, Inc. S-1 registration statement — March 20, 2026

The Federal Backstop That Most Investors Miss

X-energy is one of two companies selected by the Department of Energy in 2020 as winners of the Advanced Reactor Demonstration Program, which provides up to $1.2 billion in federal cost-share matching to develop, license, and demonstrate an operational advanced reactor and fuel facility. By December 31, 2025, the company had received approximately $438 million in ARDP reimbursements, effectively subsidizing a significant portion of its development spend.

The TRISO-X fuel fabrication facility in Oak Ridge, Tennessee — called TX-1 — received an NRC Special Nuclear Material License in February 2026 and is designed to produce enough fuel to support the first 11 Xe-100 reactors at steady state. In April 2025, TRISO-X received the first allocation of high-assay low-enriched uranium, or HALEU, from the DOE’s HALEU Availability Program, securing fuel feedstock for the initial core loads of the Texas project. These are not minor milestones. The fuel chain is the single hardest logistical problem in advanced nuclear deployment, and X-energy has meaningfully de-risked it relative to most competitors.

The Financials: What the S-1 Shows

This is where the story requires discipline. X-energy reported $109 million in revenue and grant income for 2025, down 9% from the prior year. That decline matters. Stripping out grants, revenue from operations was approximately $94 million, against a net loss of roughly $390 million, per Bloomberg’s reporting on the S-1. The prior year’s net loss was $126 million on $84 million in revenue. The loss nearly tripled year-over-year as development spending accelerated.

The company employs roughly 916 people as of March 2026. It is in what Seeking Alpha’s IPO Edge analysis called “a high cash burn phase, with sharply rising losses, heavily negative cash flow.” That is accurate and investors should sit with it. Pre-commercial nuclear development at this stage of licensing and construction activity is extraordinarily capital-intensive. The IPO proceeds are intended to fund continued reactor licensing, construction activities, and supply chain expansion — not to make the company profitable in the near term. Profitability is a post-2030 story contingent on the Seadrift plant coming online.

KEY RISKS FOR INVESTORS

Regulatory timeline risk. The NRC’s 18-month review of the Seadrift construction permit runs through approximately late 2026. Any extension or public comment complication pushes the construction start and, consequently, first revenue from operations.

Technology execution risk. No Xe-100 has ever been built. The reactor design is sound on paper and well-advanced in licensing, but first-of-a-kind construction always surfaces cost and schedule surprises.

HALEU supply chain risk. High-assay low-enriched uranium has historically been sourced primarily from Russia. X-energy received its first domestic HALEU allocation in April 2025, but fuel supply chain security for a scaled fleet remains a geopolitical dependency.

Valuation uncertainty. Share price and total raise are undetermined. Investors are being asked to price a company whose revenue case depends entirely on commercial deployments that won’t generate meaningful income until the early 2030s.

Why the Power Demand Argument Is Real

The thesis underneath X-energy’s IPO is not difficult to understand. Goldman Sachs estimates that data center electricity demand could rise 160% by 2030. The Federal Energy Regulatory Commission projects U.S. data center electricity demand will climb from 19 gigawatts in 2023 to 35 gigawatts in 2030. Goldman Sachs has further estimated that 85 to 90 gigawatts of new nuclear capacity would be needed to meet all projected data center power demand growth by 2030 — and less than 10% of that will be available globally by then.

That gap is the market X-energy is building into. It is not a manufactured thesis. Microsoft signed a 20-year power purchase agreement to restart Three Mile Island. Google contracted with Kairos Power for a fleet of SMRs targeting 2030 deployments. Meta issued an RFP for one to four gigawatts of new nuclear generation. These are not exploratory conversations. They are signed agreements and capital commitments by companies with unlimited energy budgets and existential reasons to secure reliable baseload power. X-energy has two of those companies as equity investors and development partners.

What This IPO Is Really Selling

X-energy is not a bet on current earnings. It is a bet on a structural energy transition that is already underway, on a technology that has cleared more regulatory hurdles than any other advanced reactor design in the U.S., and on a management team that includes a former Deputy Secretary of Energy and a CFO who ran corporate development for Amazon’s climate investment portfolio. Those are credentialed people with institutional backing executing against a plan that has federal cost-share and Fortune 500 anchor partnerships behind it.

The risk is time. Nuclear projects that slip slip expensively. If the Seadrift NRC review extends, if TX-1 fuel production encounters delays, or if HALEU supply chain constraints re-emerge, the cash burn against a thin revenue base will pressure the company in ways that IPO proceeds alone may not fully cushion. Investors who understand that this is an infrastructure company in the licensing phase — not a software company with a growth multiple — are the appropriate audience for this offering.

The roadshow timeline is not yet public. Based on standard SEC review periods and the March 20 filing date, trading could begin as early as May 2026. The amended S-1, which will carry the price range, is the next document to watch.

WHAT TO WATCH NEXT

The amended S-1 with pricing terms. The first filing carried no share count or price range. The amended filing will be the first hard data point for valuation analysis and the clearest signal of where institutional demand is pricing the offering.
NRC review progress on the Seadrift permit. The 18-month review clock started in May 2025, putting a decision around late 2026. Any NRC request for additional information or public hearing triggers would extend that timeline and reprice the risk of first commercial operations slipping past 2030.
TX-1 fuel facility milestones. The Oak Ridge TRISO-X facility received its NRC license in February 2026. Watch for announcements of commercial fuel production beginning — that milestone converts the fuel supply thesis from regulatory to operational.
Comparable nuclear IPO performance. X-energy will price into a market that has seen NuScale Power face cost overruns and schedule delays. How institutional investors price the sector premium or discount relative to NuScale’s public market experience will shape the deal’s reception.
Tech company nuclear procurement announcements. Any new hyperscaler power purchase agreement or SMR development partnership announced in the pre-roadshow window will serve as real-time demand validation for X-energy’s commercial thesis.

$300 Billion in 90 Days: Why the AI Funding Boom Is Different This Time

Stacking Trades — Thu, 02 Apr 2026 00:29:54 +0000

The first quarter of 2026 produced a number that no historical framework was built to absorb. Investors deployed $297 billion into roughly 6,000 startups globally between January and March, a sum that exceeds every full-year venture total before 2018 and equals nearly 70% of all capital deployed across the entirety of 2025. Call it what you want. The scale is without precedent.

The question sophisticated investors are asking is not whether the number is real. It is. The question is whether the capital concentration underlying it reflects rational conviction or the architecture of a crash in slow motion. The evidence, weighed carefully, tilts toward the former.

$297B

81%

$19B

Global VC deployed
Q1 2026 (Crunchbase)

Share flowing to
AI companies

Anthropic annualized
run rate, March 2026

Four Companies, $186 Billion, and the Case for Concentration

The headline figure is dominated by four rounds. OpenAI closed a $122 billion raise that pushed its valuation to $852 billion. Anthropic added $30 billion at a $380 billion post-money valuation. Elon Musk’s xAI pulled in $20 billion. Waymo secured $16 billion. Those four deals alone totaled $188 billion, or roughly 64% of all global venture activity for the quarter.

Critics have reached immediately for the word “concentration” as a synonym for fragility. That reading misses the structural logic. The AI buildout is not a software story in the old sense. It is simultaneously a compute story, an infrastructure story, and an energy story. Capital requirements of this magnitude are not a sign of speculation; they reflect the genuine cost of building systems that will underpin the next generation of enterprise software. These are not companies with a pitch deck and a domain name.

Early-stage funding grew 41% year-over-year, to $41.3 billion across 1,800 deals. Seed funding rose 31%. The ecosystem is not hollowing out below the megarounds. Every stage grew. The money at the top is not crowding out the next generation; it is pulling it forward.

The Revenue Underneath the Valuations

The most durable rebuttal to bubble-talk is not the funding numbers. It is the revenue curve. Anthropic’s annualized run rate hit $14 billion in February 2026, confirmed in the company’s own Series G disclosure. By March, that figure had crossed $19 billion, up from $9 billion at the end of 2025 and $1 billion in December 2024. That is not a valuation story. That is a sales story.

The growth rate has no precedent in enterprise software. More than 500 of Anthropic’s customers now spend over $1 million annually, and eight of the Fortune 10 are active Claude users. Anthropic now captures over 73% of spending from companies buying AI tools for the first time, a figure that stood at 40% just ten weeks prior. Enterprise adoption is not a lagging indicator here. It is the leading one.

“AI in 2026 is a story of capital building in silicon and steel, and in genuine utility. The doomsayers are making this millennium’s most expensive error: mistaking the most significant digital infrastructure overhaul for a speculative fantasy.”
— Sify Technologies analysis, January 2026

OpenAI, meanwhile, is running at an annualized revenue rate of approximately $25 billion, according to reporting from The Information. Epoch AI’s analysis projects that Anthropic could surpass OpenAI in annualized revenue by mid-2026 if current trajectories hold, a crossover that would represent one of the fastest competitive market shifts in enterprise software history.

This Is Not the Dot-Com Bubble. The Infrastructure Comparison Matters.

The dot-com analogy is the most repeated frame in AI skeptic commentary, and it is also the least instructive one. Pets.com raised money on a business model with no path to unit economics. Webvan tried to deliver groceries before the logistics existed. The companies absorbing the majority of Q1’s capital are generating tens of billions of dollars in annual revenue from paying enterprise customers. The comparison collapses on contact with the facts.

What the dot-com era did share with today is that transformative infrastructure buildouts look irrationally expensive until they don’t. Every fiber optic cable laid in the late 1990s that made the broadband internet possible looked like waste at the time. The waste was real. So was the infrastructure. The companies that built the physical layer of AI — data centers, specialized chips, power infrastructure — are creating assets that will be productive for decades regardless of which model wins the next benchmark.

Goldman Sachs estimates that AI capex now represents approximately 0.8% of GDP, still nearly half the 1.5% of GDP reached during comparable tech booms over the past 150 years. The cycle, by that measure, has more room to run before the historical warning signs activate.

Where the Legitimate Risks Sit

Acknowledging the bull case does not require pretending there are no risks. There are, and serious investors should hold both simultaneously.

The venture concentration is real. If three companies account for the lion’s share of a quarter’s capital, the health of that quarter is hostage to the health of those three companies. A major regulatory shock, a security incident, or a model capability plateau at any of the frontier labs would register not as a sectoral speed bump but as a funding crisis. The interdependencies run deep.

The exit environment remains thin. Only four US venture-backed companies exited above $1 billion in Q1 via IPO, against a backdrop of surging valuations and enormous unrealized paper gains sitting on LP balance sheets. The IPO market has not yet reopened at a scale commensurate with the private market’s ambitions. That gap will not close on its own.

Morgan Stanley’s strategists, in a February 2026 note, said plainly that “the AI buildout has become so large — and so well understood — that it no longer supports paying any price for the companies driving it.” That is a useful discipline for public market investors. It is also, notably, a statement about pricing discipline, not about whether the underlying technology or its commercial potential is real.

Unlike Any Previous Cycle

What distinguishes the current moment from prior technology booms is not just the dollar figures. It is the physical dimension of the investment. Capital is flowing into autonomous vehicles, robotics, and manufacturing alongside model development. The AI boom is simultaneously a software cycle and an industrial cycle, with capital requirements that match. That duality is new.

The secondary market for private stakes in AI companies is also maturing rapidly, with $240 billion in secondary transaction volume in 2025 and projections approaching $300 billion for 2026. Liquidity mechanisms that did not exist two years ago are creating pricing signals and investor discipline that help distinguish genuine value from hype at the asset level. The private market is developing its own immune system.

Q1 2026 will be studied for a long time. The more interesting question is whether the revenue growth that underlies these valuations continues to outpace the capital being deployed into them. So far, the answer is yes. That is the number to watch.

WHAT TO WATCH NEXT

SpaceX’s public S-1 filing. The confidential filing is done. The public prospectus — expected roughly eight weeks later — will be the first hard look at the revenue and cost structure of the largest IPO ever attempted, and will test whether public markets can absorb AI-era private valuations at scale.
OpenAI and Anthropic revenue pacing through Q2. Epoch AI projects a potential revenue crossover between the two companies by mid-2026. Whether that materializes — and how publicly either company discloses it — will be a significant signal for private market pricing across the sector.
The IPO backlog. Only four US companies exited above $1 billion via IPO in Q1. The pressure on GPs to return capital to LPs is intensifying. Watch whether the SpaceX listing, if it proceeds, creates a window for other high-profile AI-adjacent companies to follow.
Enterprise AI spend in Q1 earnings. The hyperscalers report Q1 earnings in late April. Any softness in cloud revenue or AI-related guidance will test whether enterprise adoption is accelerating as fast as the private market assumes.
Secondary market pricing. As dry powder in dedicated secondary strategies approaches $327 billion, pricing premiums or discounts on AI company stakes will serve as the most honest real-time valuation signal available outside of a formal IPO.

The $75 Million Wall: Regulation A+ Is Running Out of Room

Stacking Trades — Wed, 01 Apr 2026 23:06:04 +0000

When the SEC raised the Regulation A+ Tier 2 offering cap to $75 million in late 2020, the intent was straightforward: give growing companies more room to raise public capital without the full burden of a registered IPO. Five years on, the cap looks less like a runway and more like a hard stop — one that successful issuers keep hitting, and that many others are avoiding entirely because the economics no longer make sense at that ceiling.

The numbers are stark. Reg A+ issuers raised approximately $1.85 billion in 2022 across roughly 307 qualified offerings. By 2024, that figure had collapsed to $896 million across just 102 qualified offerings — a 52% drop in proceeds and a 67% drop in deal count from peak. Meanwhile, Regulation D — available only to accredited and institutional investors — raised approximately $2.15 trillion in 2024 alone.

That gap is not just a market-structure curiosity. It represents a capital formation bottleneck that Congress and the SEC are now actively working to address, with competing proposals on the table and a Senate review process that could reshape the market’s architecture for the next decade.

Why the Cap Creates a Trap, Not Just a Limit

To understand the problem, you have to understand what happens when a Reg A+ issuer approaches $75 million. Issuers conducting continuous offerings who want to raise beyond the cap in a rolling 12-month period must file either a post-qualification amendment (PQA) or an entirely new offering statement on Form 1-A. Both options require SEC review and re-qualification before sales can resume.

The problem is that the PQA process creates serious operational friction precisely when issuers are most active. Goodwin’s analysis walks through the mechanics: if an issuer raises $5 million in each of the first two months of a continuous offering and then sells the remaining $65 million over the next ten months, it can only file a PQA at month 13 to qualify the $5 million that has freed up. If it wants to keep selling at pace, it has to file consecutive PQAs each month — each requiring legal opinions, auditor consents, and SEC review. The cost per amendment can run into the tens of thousands of dollars before professional fees are counted.

The SEC’s own 2018 rulemaking estimated that issuers spend approximately 731 hours preparing and filing a single Form 1-A offering statement. That burden compounds with every PQA. The message that sends to scaling companies: once you’re close to $75 million, the marginal cost of staying in Reg A+ starts to rival the cost of a full registered offering.

“The top reason cited by Goodwin clients for not utilizing Regulation A is that the dollar limit is too small given the amount of time and money it would take to launch an offering.”
— Goodwin Law, November 2026

Congress Steps In With a Double

The legislative response arrived in December 2025. Rep. Marlin Stutzman introduced the Regulation A+ Improvement Act of 2025 (H.R. 6541), which would double the Tier 2 cap from $75 million to $150 million annually and add an inflation-adjustment mechanism tied to the Consumer Price Index, recalibrated every five years. The House Financial Services Committee advanced the bill 28-23 on December 17, 2025, along party lines.

The bill was folded into the broader INVEST Act package (H.R. 3383), which passed the full House 302-123 on December 11, 2025. The bill was received in the Senate on December 15 and referred to the Senate Banking, Housing, and Urban Affairs Committee. As of early February 2026, no Senate action had been taken, with timing and the possibility of amendment remaining unclear.

The bipartisan House margin is notable — it signals genuine political demand for reform — but the Senate is a different committee. Senate Banking Chair Tim Scott has championed capital formation legislation in prior sessions, which proponents cite as a favorable signal. Whether the Reg A+ provisions survive Senate markup intact is a separate question.

The Cap Debate Is Bigger Than $150 Million

Among practitioners, $150 million is considered a floor, not a destination. Goodwin has formally called for raising the cap to at least $300 million. Former SEC Chair Jay Clayton suggested the appropriate number might be $1 billion to $2 billion. The SEC’s own small business forum considered $150 million as a reference point as recently as its 44th annual report.

These positions reflect a structural argument: Reg A+ was conceived as a bridge between the fully private markets of Regulation D and a registered public offering. At $75 million — or even $150 million — the bridge doesn’t reach far enough for companies that have found genuine product-market fit and need institutional-scale capital to continue growing before they’re ready for an S-1.

Goodwin’s November 2025 white paper also flagged a mechanic that has been largely overlooked: the annual PQA filing requirement creates a compounding burden for successful continuous-offering issuers that is entirely disconnected from investor protection. Their proposal would allow issuers to bank an additional $75 million of qualification capacity into the annual required PQA — so that active issuers don’t have to file separately every time the rolling 12-month window frees up capacity.

The Sector Concentration Problem Nobody Wants to Talk About

The strongest counterargument to expanding Reg A+ isn’t that the cap is set correctly — it’s that the market’s composition raises its own concerns. SEC DERA data found that financial sector issuers accounted for roughly 46% of aggregate financing sought under Reg A+ and an even larger share — approximately 64% — of actual reported proceeds. Real estate issuers, REITs, holding companies, and non-depository financial institutions dominate the top of the leaderboard.

The House Financial Services Committee minority noted this directly in the H.R. 6541 committee report: a $150 million ceiling would make the framework more attractive at scale for exactly these financial-sector vehicles, increasing retail exposure to illiquid and complex products. The concern is less about operating companies issuing Reg A+ shares and more about structured finance vehicles using the framework to access a retail investor base that may not have the tools to evaluate what they’re buying.

This is a real tension. Reg A+ was built around the idea of investor access — allowing unaccredited investors to participate in private-market-style offerings subject to investment limits. If the framework’s primary beneficiaries at scale are shadow finance vehicles and real estate vehicles rather than operating startups, lifting the cap may simply make an existing concentration worse before it gets better.

What the Decline in Filings Actually Signals

The 67% drop in Reg A+ offering volumes from peak is doing several things at once, and conflating them produces the wrong diagnosis. Part of the decline reflects the normalization of pandemic-era capital markets activity — Reg CF and Reg A+ both surged in 2020-2022 on the back of retail investor enthusiasm and low rates. The hangover was inevitable.

But the structural argument holds even after discounting the cycle. The drop in Reg A+ filings significantly exceeded the drop in broader capital market activity, suggesting that the framework’s structural costs — compliance overhead, cap constraints, the PQA cycle — are driving away precisely the mid-size issuers the tool was designed to serve. Companies that could realistically raise $50 million to $100 million and graduate toward a public listing are doing Regulation D rounds instead, limiting their investor base to accredited investors and forgoing the retail capital access that was Reg A+’s original value proposition.

That outcome benefits nobody the reform was designed to help: not the retail investors who are locked out of the earlier stages of company growth, and not the companies that could build a broader shareholder base as a stepping stone to eventual public markets.

WHAT TO WATCH NEXT

Senate Banking Committee action on the INVEST Act — specifically whether the Reg A+ Improvement Act provisions survive markup intact, get amended to a different cap level, or get stripped from the package entirely.
Whether the SEC under Chair Paul Atkins moves independently to raise the cap via rulemaking, as the agency has authority to do without Congress, and whether that would pre-empt or complement legislative action.
DERA’s next exempt offering data release — watch whether 2025 Reg A+ volumes continue declining or stabilize in anticipation of legislative clarity, and whether the financial sector’s share of proceeds grows or contracts.
Post-qualification amendment filing volume on EDGAR for active continuous-offering issuers: PQA frequency is the clearest leading indicator of how close issuers are to the cap in real time.
Any formal SEC rulemaking on PQA review timelines and automatic qualification of non-material amendments, which Goodwin explicitly proposed and which could provide meaningful relief even without a cap increase.