Space Exploration Technologies Corp. (Nasdaq: SPCX) rose 23% last week, closing Friday at $133.11. That snapped four straight losing weeks and brought the stock back to within 1.4% of its $135 IPO price.

A big number is driving the excitement. A SemiAnalysis research note argued SpaceX could spend $300 billion to $500 billion in capital expenditures, or capex — the money spent building data centers, buying chips, and supplying power — in 2027 alone. That’s a modeled estimate, not a disclosed spending plan. Two pieces of that story still need signatures. One from a company that might help finance the build. One from a company that might buy most of what gets built. Neither agreement has been publicly disclosed.

Here’s the frame for the whole issue. Real, modeled, and speculative are three different things. This week’s rally blends all three together. What’s real: SpaceX’s capacity ambition, its exclusive commitment to NVIDIA Corp. (Nasdaq: NVDA), and disclosed compute deals with Alphabet Inc.’s (Nasdaq: GOOGL) Google and Anthropic. What’s modeled: the $300 billion to $500 billion 2027 capex figure and the revenue math behind it. What’s still speculative: NVIDIA financing that build, and Microsoft Corp. (Nasdaq: MSFT) becoming the largest buyer of the output.

Thursday’s 6.1% gain came even as more than 900 million insider shares became eligible to trade for the first time. That doesn’t prove index funds started buying, or that short covering explains the whole move. It does show the first unlock session cleared without the flood of selling investors feared — not that insiders are done selling, but that the market absorbed the first day of dramatically greater supply better than expected.

Friday’s 16% surge had several sources, not one. The earnings beat looked better once investors had time to digest it. The lockup overhang seemed less damaging than feared. Analysts raised forecasts, and equity research firm Argus upgraded SpaceX to Buy that same day. The broader market rallied too, as a mild jobs report eased rate-hike worries and the semiconductor index climbed 2.6%. SemiAnalysis published its aggressive capacity model that same morning.

The research note added fuel. It wasn’t the only match.

SemiAnalysis is an independent semiconductor and AI-infrastructure research firm. Its model: SpaceX adds 6 to 8 gigawatts of incremental AI computing capacity during 2027, taking total capacity toward roughly 10 gigawatts by year-end. A gigawatt is one billion watts of power — a utility-scale amount of electricity. SemiAnalysis models the cost at roughly $50 billion per gigawatt. The six-to-eight-gigawatt base case implies roughly $300 billion to $400 billion of capex. The $500 billion upper end only applies if the incremental build reaches closer to 10 gigawatts. That’s SemiAnalysis’s model. It is not spending guidance SpaceX itself has issued.

SpaceX doesn’t have $500 billion of cash sitting in the bank. It also wouldn’t need the whole amount on day one. A buildout this size would likely draw on several sources at once. Existing cash. Operating cash flow. Customer payments. Debt. Project financing. And, SemiAnalysis argues, possibly supplier support too. SpaceX ended June with roughly $100 billion of cash and marketable securities, plus $47.5 billion of backlog — contracted business not yet recognized as revenue. This isn’t currently an inability-to-pay story. It’s a funding-mix question. How much needs outside financing? Who supplies it? What do they get back in return?

SemiAnalysis believes SpaceX may lean on vendor financing to lower that upfront cash need. Vendor financing is when a supplier helps a customer pay for the supplier’s own product. Here, the supplier would be NVIDIA, and the product is the processors inside SpaceX’s AI infrastructure. That financing hasn’t been announced. SemiAnalysis infers NVIDIA may provide that support. It also infers SpaceX’s decision to go NVIDIA-exclusive could be part of the bargain. That’s a plausible theory. It is not a signed term sheet.

Vendor financing wouldn’t eliminate SpaceX’s upfront funding need. It would change who supplies the capital — and what NVIDIA receives in return. That distinction is the credibility firewall for this whole issue.

The financing theory answers who might fund the build. SemiAnalysis’s Microsoft thesis answers who buys the output.

The research firm believes Microsoft Corp. (Nasdaq: MSFT) could become SpaceX’s largest compute customer, or “offtaker” — the company that buys and uses the capacity SpaceX builds. Its model estimates Microsoft could generate more than $100 million of annual revenue per megawatt running OpenAI’s models across Azure, Foundry, and Copilot. It also argues SpaceX can bring large blocks of capacity online in three to five months, far faster than a conventional data center project.

Those are model outputs and analyst conclusions. Microsoft has not disclosed a SpaceX compute contract. Until it does, Microsoft isn’t the receipt. It’s the next validation event.

SpaceX doesn’t need Microsoft to prove real demand exists. It already has disclosed contracts, confirmed in SEC filings. Anthropic agreed to pay $1.25 billion a month for compute tied to roughly 325,000 NVIDIA GPUs, through May 2029. Google agreed to pay $920 million a month, from October 2026 through June 2029, for compute tied to roughly 110,000 NVIDIA GPUs. Both deals include a 90-day termination window — Anthropic’s opens after an initial three-month period, Google’s after December 31, 2026.

Those are real customers and real contracted payments. They are not equivalent to fully non-cancellable, 15-year take-or-pay commitments. SpaceX has proved customers will rent its compute. It hasn’t yet proved every gigawatt in the 2027 model has a durable buyer. Microsoft would strengthen that case considerably. It’s not the first customer, and it isn’t a confirmed one yet.

Barron’s, citing JPMorgan tracking data, estimates retail investors have lost roughly $4.5 billion combined since the IPO, buying at an average of around $150 a share — well above Friday’s $133.11 close. Reuters also reported retail investors turned net sellers for the first time since the IPO Friday, though the modest size of that net sale next to their busiest buying day looks more like fatigue than capitulation. A stock can have its best week ever while much of its own shareholder base is still waiting to break even.

SpaceX traced a full boom-bust-rebound arc in its first two months as a public company. Friday may mark a tactical low. It hasn’t proven the final bottom is in.

The rest of the space-stock group rallied too, but not as a clean SpaceX sympathy trade. Rocket Lab USA Inc. (Nasdaq: RKLB) gained 9.5% Friday — strong, though smaller than SpaceX’s roughly 16%. Rocket Lab had catalysts of its own: its 92nd Electron rocket launch and a newly announced $397 million U.S. Space Force contract. AST SpaceMobile Inc. (Nasdaq: ASTS) rose 6.8% days after launching three new BlueBird satellites on a SpaceX Falcon 9 — a real SpaceX relationship, but a launch contract, not exposure to the AI-datacenter buildout. Redwire Corp. (NYSE: RDW) jumped 14.9% on its own record quarterly revenue and backlog, plus a newly announced commercial mission aboard a SpaceX spacecraft. Intuitive Machines Inc. (Nasdaq: LUNR) added 9.8% after entering the week with catalysts of its own, including its Goonhilly/COMSAT acquisition and new missile-tracking work through L3Harris.

Iridium Communications Inc. (Nasdaq: IRDM) and Globalstar Inc. (Nasdaq: GSAT) barely moved. EchoStar Corp. (Nasdaq: ECHO) has the sector’s clearest disclosed SpaceX tie — it moved the spectrum assets it’s selling SpaceX into a trust in May, with the final acquisition closing still targeted for 2027 — yet its stock fell 5% Friday anyway. A direct SpaceX relationship didn’t guarantee participation in the SpaceX rally.

SpaceX restored risk appetite across the space-stock group. Several of these companies also had real catalysts of their own. SpaceX’s record IPO probably drained some speculative capital from the rest of the complex, as investors raised cash to participate — though there’s no stock-by-stock flow data proving that. A bigger AI-and-space infrastructure market could someday expand the opportunity set for launch providers and component suppliers, but SpaceX builds much of its own stack, and no public peer has disclosed revenue tied directly to its AI buildout. A stock rallying alongside SpaceX because sentiment turned is not the same as a stock rallying because SpaceX’s AI-infrastructure demand flows through to its own revenue.

$RKLB ( ▼ 5.91% ) had earnings last night and is down almost 9% this morning.

NVIDIA Corp. (Nasdaq: NVDA). The exclusive commitment tied to SpaceX’s Vera Rubin buildout is real. It’s one of the clearest long-term demand signals NVIDIA has received this year. Risk: any vendor-financing arrangement remains unconfirmed. Financing a customer’s purchase is economically different from receiving an independently funded order.

Space Exploration Technologies Corp. (Nasdaq: SPCX). The stock showed that investors are willing to reward a credible path to much larger AI revenue. It also showed that the first unlock session could clear without the feared collapse. Risk: the capex figure, the NVIDIA support, and the Microsoft offtake are all modeled or inferred. None of it is signed and disclosed yet, and another scheduled share release is still ahead later this month.

Rocket Lab USA Inc. (Nasdaq: RKLB). Benefited from the sector’s restored risk appetite, plus real catalysts of its own — its 92nd Electron launch and a new $397 million Space Force contract. Risk: no disclosed commercial tie to SpaceX’s AI-datacenter buildout specifically.

Microsoft Corp. (Nasdaq: MSFT). Potentially the largest beneficiary and offtaker under the SemiAnalysis model. That model shows a path to accelerating Azure’s growth rate. What’s missing: a disclosed Microsoft-SpaceX compute agreement.

CoreWeave Inc. (Nasdaq: CRWV) and Nebius Group N.V. (Nasdaq: NBIS). Wall Street research firm Bernstein warns SpaceX could get priority over both neocloud providers — smaller cloud companies that rent out AI chips — in future NVIDIA allocations. That’s a legitimate analyst concern, not evidence of lost deliveries yet.

Advanced Micro Devices Inc. (Nasdaq: AMD). SpaceX’s exclusive NVIDIA decision excludes AMD from this phase of the buildout. That’s a real loss, though it doesn’t mean AMD is shut out of every future SpaceX deployment. Risk to the bearish view: AMD still retains major customers. One buyer doesn’t decide the whole chip market.

We continue to own SPCX in 3 of our ETFs: $SPCI ( ▼ 4.02% ) $HALX ( ▲ 0.36% ) $UFOD ( ▼ 2.62% ) . We are not adding, or taking off exposure.

The operating evidence — Anthropic, Google, and the NVIDIA exclusivity — is real. The $300 billion to $500 billion buildout is modeled. The NVIDIA financing architecture is speculative. I don’t trade an analyst model as though it were a signed term sheet.

I want to see three things before I get more constructive on SPCX itself. A disclosed financing structure. A named large-scale offtaker beyond the customers already public. And evidence the economics still work once interest and depreciation eat into them. CoreWeave reports Tuesday after the close (5:00 p.m. ET call); Nebius reports Wednesday before the open (8:00 a.m. ET call). I’m listening for processor allocations, deployment delays, and any change in access to next-generation NVIDIA systems.

The financing signature. I get more constructive if SpaceX or NVIDIA discloses the actual financing structure — the amount, form, term, pricing, collateral, and recourse. Vendor financing doesn’t have to mean a conventional loan; it could be trade credit, leases, guarantees, or deferred payments. I get more cautious if disclosed terms show NVIDIA funding a material share of purchases that become its own revenue, without strong collateral, meaningful recourse, or evidence of independently funded customer demand.

The customer signature. I get more bullish on the whole SpaceX AI story if Microsoft — or another large buyer beyond Google and Anthropic — signs a confirmed, disclosed contract large enough to support the next stage of the capacity plan. Right now Microsoft is still a model, not a customer.

The neocloud squeeze. I get more concerned about CoreWeave and Nebius if this week’s earnings show actual chip-delivery delays. Analyst speculation about future allocation alone won’t move me.

The sector read-through. I get more convinced this is durable if any space-stock peer discloses real revenue tied to SpaceX’s AI-infrastructure buildout — not just shared sentiment.

SpaceX has shown it can land real customers and execute an ambitious infrastructure buildout. SemiAnalysis modeled what that execution speed could mean at data-center scale. What the next stage still needs are two public signatures: one from a financier willing to fund enough of the build, and one from an additional customer large enough to underwrite the next wave of capacity beyond Google and Anthropic.

The rest of the space sector shared the rally. It has not yet shown that it shares the economics. The rally went sector-wide. The receipts did not.

NVIDIA may become that financier. Microsoft may become that customer. Neither agreement has been publicly disclosed. The $500 billion buildout isn’t nonsense. It’s a model waiting for contracts.

I haven’t talked about Bitcoin here in a long time, will have to rectify that soon. In the meantime here are some of my thoughts on why it belongs in a portfolio……

The news. The fine print is the news. Nvidia sold the market a $500 billion “third-party capital” platform for AI infrastructure, with Apollo, BlackRock, Blackstone, Brookfield, Goldman and KKR lined up to finance compute. Reuters says the company disclosed no commitments, no financial terms and no detail on how the number fits with existing deals. The FT then found the part that matters: Jensen Huang said Nvidia may provide residual-value support for up to 25% of projects. That is not nothing. It means Nvidia is not only selling the picks and shovels. In some cases, it may be helping protect the financing value of the mine. Demand is real. But when a supplier needs Wall Street to package the customer’s buying power, the story has moved from product cycle to credit cycle. (Reuters — Nvidia financing, FT — Nvidia becomes the bank of AI, MarketWatch — Nvidia financing platform)

The noise. The noise is that this proves either AI is fake or Nvidia is invincible. Both are lazy. AI compute demand is enormous. That is why the financing exists. But the market is pretending “third-party capital” makes the risk disappear. It does not. It moves the risk into asset-backed structures, private credit, residual values, utilization assumptions and end-customer solvency. At the same time, oil is back near $88 because U.S.-Iran talks over Hormuz hit an impasse, and money markets are back to a roughly 50/50 chance of a September hike. Tomorrow’s CPI will not capture the latest oil move. That matters. The AI buildout needs cheap capital, stable power and customers who can monetize compute. Warsh, oil and credit are all pushing the other way. (Reuters — oil and Fed odds, Barron’s — inflation worries, Barron’s — Bank of Nvidia)

The dumb advice. The dumb advice is treating “compute is revenue” as an investment conclusion. Compute is not revenue. Compute is capacity. Revenue shows up only if customers sell enough tokens, cloud usage, ads, subscriptions or workflow automation to pay for it. Goldman’s David Solomon said he is excited to create a credit market backed by Nvidia compute. Of course he is. Wall Street loves turning hot demand into a tradable asset class. That may be good for Goldman. It does not automatically make the borrower a good equity or the chip order a high-return investment. Intel just fell after announcing a $15 billion stock sale to fund AI demand. Alphabet and Tesla already showed negative free cash flow while ramping AI spending. The issue is not whether AI works. The issue is whether the marginal dollar of financed compute earns more than it costs. (MarketWatch — Solomon quote and Nvidia platform, AP — Intel stock sale, Barron’s — circular financing concerns)

Concrete takeaways.

The HEAT (Hedge, Edge, Asymmetry and Theme) Formula is designed to empower investors to spot opportunities, think independently, make smarter (often contrarian) moves, and build real wealth.

The views and opinions expressed herein are those of the Chief Executive Officer and Portfolio Manager for Tuttle Capital Management (TCM) and are subject to change without notice. The data and information provided is derived from sources deemed to be reliable but we cannot guarantee its accuracy. Investing in securities is subject to risk including the possible loss of principal. Trade notifications are for informational purposes only. TCM offers fully transparent ETFs and provides trade information for all actively managed ETFs. TCM's statements are not an endorsement of any company or a recommendation to buy, sell or hold any security. Trade notification files are not provided until full trade execution at the end of a trading day. The time stamp of the email is the time of file upload and not necessarily the exact time of the trades. TCM is not a commodity trading advisor and content provided regarding commodity interests is for informational purposes only and should not be construed as a recommendation. Investment recommendations for any securities or product may be made only after a comprehensive suitability review of the investor’s financial situation.© 2026 Tuttle Capital Management, LLC (TCM). TCM is a SEC-Registered Investment Adviser. All rights reserved.