I’ve been a trader and investor for 44 years. I left Wall Street long ago—-once I understood that their obsolete advice is designed to profit them, not you.
Today, my firm manages around $5 billion in ETFs, and I don’t answer to anybody. I tell the truth because trying to fool investors doesn’t help them, or me.
In Daily H.E.A.T. , I show you how to Hedge against disaster, find your Edge, exploit Asymmetric opportunities, and ride major Themes before Wall Street catches on.
Table of Contents
H.E.A.T.
Up to +10% META intraday move on Bloomberg report; Reuters confirmed >7% early trading | $125–145B Meta 2026 capex guidance (up $10B from prior, April 2026 earnings) | >$700B Expected Big Tech AI infrastructure spend globally in 2026 (Reuters) | $35B Meta–CoreWeave total committed spend ($14.2B + $21B deals, 2025–2026) |
Sources: Meta Q1 2026 earnings call; Bloomberg (July 1, 2026); Reuters (July 1, 2026; April 2026 CoreWeave deal)
On July 1, 2026, Bloomberg reported what Meta's shareholders had been waiting two years to hear: the $125–$145 billion AI infrastructure buildout is not just a cost. It might be a business.
Meta shares rose as much as 10% intraday. Reuters confirmed the move at greater than 7% in early trading. That is not a routine reaction to a product announcement — because there was no product announcement. Bloomberg cited people familiar with the plans; Meta declined to comment; Reuters could not independently verify the report. What moved the stock was not a launch. It was a narrative shift.
For two years, every Meta earnings call has felt like a hostage negotiation with the capex line. Investors watched the number go up — $60B, $80B, $100B, now $125–$145B — and kept asking the same question: what is the return? Bloomberg's report offered one answer. The infrastructure being built to serve Meta's own AI ambitions could also be rented. The capex becomes the product.
That distinction changes everything about how you should think about Meta, about the cloud market, and about every company caught in between.
PART I: THE CAPEX PROBLEM THAT JUST BECAME A REVENUE NARRATIVE
Let's be precise about what Meta is and isn't announcing, because the investment implications live entirely in that gap.
The Bloomberg report describes an internal initiative — Meta Compute — with two distinct business lines in development. The first is a model-hosting platform analogous to AWS Bedrock: Meta runs the infrastructure, developers pay to call the models, including Meta's proprietary Muse Spark models. The second is raw compute rental analogous to CoreWeave or Lambda Labs: GPU time and data center capacity, sold by the rack.
Neither is live. Neither has a launch date. What we know for certain is: Meta has the infrastructure, the models, and the reported intention. What the market is pricing today is the probability that it executes.
Here is why that probability is non-trivial.
Meta raised its 2026 capex guidance in April to $125–$145 billion — up $10 billion — citing higher component pricing and additional data center costs. That language is not the language of a company pulling back. It is the language of a company accelerating into a constraint it can see but hasn't solved. The capacity being built is not merely an input to Meta's own products. Per the Bloomberg report, it is a potential asset — one that can be rented while the owner occupies the lobby.
SCALE CHECK Meta has major compute partnerships already in place and publicly confirmed: CoreWeave ($14.2B deal 2025, $21B expansion April 2026), Google, and Oracle. Its MTIA custom inference chip is in production deployment across Meta's own data centers. The raw ingredients for a competitive cloud offering are already running. The question is not whether Meta can build a cloud. It is whether it can sell one. |
PART II: THE COMPETITIVE MAP — WHO GETS DISRUPTED AND WHO GETS A TAILWIND
The instinct is to frame this as a war between Meta and the hyperscalers. That framing is too simple — and it misses the more investable angles.
AWS / Azure / Google Cloud: The headline threat is real but overstated in the near term. The corporate CIO who hosts production databases on AWS is not moving to Meta Compute in 2027. Enterprise switching costs are enormous, compliance infrastructure takes years to build, and Meta's brand is consumer social media — not the reassurance a CIO needs.
But AI-native developers operate on different logic. They care about model quality, latency, and price — less anchored to a 15-year enterprise relationship. Meta already has them. LLaMA has been downloaded hundreds of millions of times. A hosted API at competitive pricing is a funnel problem, not a trust problem. The AI-native developer segment is where AWS is most contestable, and it is exactly where Meta has existing distribution.
CoreWeave — the most asymmetric risk in the stack: Meta has committed $35 billion to CoreWeave across two deals. That spending validates CoreWeave's model and drives near-term revenue. But Meta Compute is structurally CoreWeave's nightmare: the largest growth customer learning to build the product it's renting. Microsoft accounted for approximately 67% of CoreWeave revenue last year; Meta is the next concentration story — in either direction.
On the day of the Bloomberg report, CoreWeave and Nebius sold off sharply. The market read Meta Compute as a substitution risk before Meta confirmed anything. The next CoreWeave earnings call is the first clean signal on whether that read was premature or prescient.
"The market spent two years punishing Meta for AI capex. It took one Bloomberg report to reprice the possibility that the capex might be right." |
PART III: THE OPEN-SOURCE TENSION — THE BUSINESS MODEL META HASN'T SOLVED
Here is where conventional analysis breaks down.
Meta's AI strategy is built on open source. LLaMA models are free to download, fine-tune, and deploy. That strategy has been enormously successful at building developer mindshare — arguably making LLaMA the most widely deployed AI model family in the world by installation count.
But open source and high-margin cloud are in structural tension. If the model is free, the only thing to sell is compute — and raw compute margins are thin. AWS Bedrock's value proposition is not just that it hosts models. It is that it integrates those models into a broader ecosystem of storage, data pipelines, compliance tools, and enterprise services that create switching costs and justify pricing power.
Meta does not yet have an AWS-style cloud ecosystem: storage, databases, IAM, compliance tooling, enterprise sales motion, support infrastructure, and multi-year CIO trust. Building it is a multi-year, multi-billion-dollar undertaking that is separate from — and arguably more difficult than — the infrastructure buildout itself.
The Muse Spark models are a partial answer. If Meta's proprietary models prove best-in-class for specific use cases — coding, multimodal reasoning, agent workflows — then hosted access to those models generates differentiation that justifies a margin premium above raw compute economics. That is the business the market is pricing today: not a GPU landlord, but a model platform.
The custom-silicon roadmap is the other piece. MTIA is real: MTIA 300 is in production, MTIA 450/500 targets inference, Broadcom is assisting on design, TSMC is fabricating. If Meta sells compute externally, lowering cost-per-token becomes a business model — not just an internal efficiency project. Qualcomm is separately confirmed as Meta's first Big Tech customer for data-center CPUs, with deployment from late 2028. Each of these is a signal that Meta is building toward owned-margin infrastructure, not rented-GPU economics.
CREDIBILITY NOTE The Bloomberg report cites 'people familiar with the plans.' Meta declined to comment. Reuters confirmed the report but noted it could not independently verify details. Muse Spark models are referenced as part of the planned offering but have not been publicly detailed by Meta. We are reflecting the weight of Bloomberg's reporting and treating the initiative as directional intelligence rather than a confirmed product launch. See Credibility Firewall below. |
WINNERS & LOSERS
COMPANY / SECTOR | VERDICT | WHY IT MATTERS | RISK |
CoreWeave | NEAR-TERM BENEFICIARY / LONG-TERM PRESSURE | Meta's $35B in committed CoreWeave spending ($14.2B + $21B, 2025–2026) validates CoreWeave's AI cloud model and drives near-term revenue. But Meta Compute is structurally CoreWeave's nightmare: the customer learns to build the product. Microsoft accounted for ~67% of CoreWeave revenue last year — Meta is the next concentration risk in either direction. | If Meta Compute scales, CoreWeave faces substitution from its largest growth customer. The next CoreWeave earnings call is the first clean read on whether Meta capacity is expanding alongside CoreWeave's or beginning to displace it. |
Nvidia | NEAR-TERM WINNER | More AI cloud capacity means more accelerator demand — every hyperscaler that enters this race writes another check to Santa Clara. Meta's buildout is GPU-intensive regardless of whether the compute is internal or external-facing. | The longer Meta turns compute into a margin business, the more incentive it has to shift inference workloads toward MTIA, AMD, or Qualcomm custom silicon. Nvidia wins the buildout; custom silicon contests the margin pool. |
Broadcom / TSMC / AMD / Qualcomm | STRATEGIC BENEFICIARY | Meta's MTIA roadmap is sourced and real: MTIA 300 already in production, MTIA 450/500 focused on inference, Broadcom assisting on design, TSMC fabricating. If Meta sells compute externally, lowering cost-per-token becomes a business model — not just an internal efficiency target. Qualcomm is separately confirmed as Meta's first Big Tech customer for data-center CPUs, with deployment from late 2028. | Custom silicon timelines are long. Nvidia's software ecosystem (CUDA) is a switching cost that ASIC vendors have not fully solved. The 2028 Qualcomm deployment is a thesis, not a 2026 catalyst. |
AWS / Amazon | PRESSURE | AWS built the cloud infrastructure playbook over 20 years. A Meta entry into model-hosting (Bedrock-style) competes directly for AI-native developer spend — the fastest-growing segment of cloud. AWS's enterprise moat is deep; the AI developer market is the contested ground. | AWS's enterprise switching costs are enormous. Meta will need years to build compliance certification, enterprise tooling, and CIO trust. The enterprise workload is not moving in 2026 or 2027. |
Microsoft Azure | WATCH LIST | Azure's value proposition is enterprise integration plus OpenAI model exclusivity. Meta's open-source LLaMA ecosystem targets a different developer audience — but a hosted Muse Spark API blurs that line and competes for AI developer mindshare directly. | Microsoft has a $13B head start with OpenAI infrastructure and deep enterprise penetration. If Meta's proprietary models prove competitive, Azure loses developer mindshare in AI-native workloads at the margin. |
Uncontracted GPU-Rental Clouds (spot capacity providers) | PRESSURE | Meta entering raw compute with $125B+ in capex is a direct price threat to commodity GPU rental — Lambda Labs, Vast.ai, and other uncontracted spot providers. The pressure is on undifferentiated capacity without long-term contracts, unique power access, or specialized tooling. | Contracted neoclouds with anchor customers and power infrastructure remain differentiated. This is not a death sentence for every neocloud — it is an accelerant on the commoditization already underway in uncontracted spot markets. |
Meta Shareholders | STRATEGIC INFLECTION | The market spent two years punishing Meta's capex. A cloud monetization path converts the liability narrative into an optionality narrative. The intraday move is the market repricing that probability — not a product launch, but a possibility it previously assigned near-zero value. | Execution risk is enormous. Building a cloud business is not like building a social app. Enterprise sales, uptime SLAs, compliance, developer tooling, and ecosystem stickiness take years to mature — and Meta starts with none of the CIO trust that makes AWS deals sticky. |
PRESSURE POINTS
PRESSURE POINT | WHAT TO WATCH | TIME HORIZON |
Meta Compute Launch Announcement | Any official product announcement — whether model-hosting (Bedrock-style) or raw capacity (CoreWeave-style) — defines the competitive perimeter and moves stocks in cloud and AI infrastructure. Meta declined to comment on the Bloomberg report; an official launch changes the information environment entirely. | Q3–Q4 2026 |
Capex Guidance Revision | Meta's Q2 2026 earnings call (July 2026) will be scrutinized for any upward revision to the $125B–$145B range and for any language around external monetization. A single sentence about revenue potential on that call reprices the capex narrative. | July 2026 |
CoreWeave Contract Signaling | CoreWeave's next earnings call and filings contain the clearest signal on whether Meta is building alongside CoreWeave's capacity or beginning to substitute for it. Volume expansion = validation. Renegotiation or reduction = Meta Compute is accelerating. | Q3 2026 |
Cloud Control-Plane Buildout | IAM, billing, observability, compliance tooling, enterprise SLAs, support, and data-residency controls — the infrastructure layer that makes cloud services sticky. Without it, Meta is a GPU landlord, not a hyperscaler. Watch for developer tooling announcements, enterprise partnership filings, and compliance certifications. | 2026–2028 |
External Compute Gross Margin | If Meta launches raw compute and margins look like commodity GPU rental, the thesis is less valuable. The real upside comes from hosted models, API routing, proprietary services, and tooling — the margin layers above raw infrastructure. First product launch disclosure will be the tell. | First product launch + first earnings disclosure |
Developer Adoption of Hosted Models | If Meta opens hosted access to Muse Spark and developers adopt at scale, it competes directly with AWS Bedrock, Azure AI Studio, and Google Vertex AI — the AI-native developer segment where brand trust matters less than model quality and price. | 12–18 months |
Export Control & Geopolitical Overhang | Any new restrictions on GPU exports or data localization rules in the EU or Asia could constrain Meta's ability to sell compute internationally — the margin-rich portion of the cloud thesis. | Ongoing — event-driven |
CREDIBILITY FIREWALL
COMPANY-DISCLOSED | SELL-SIDE & REPORTED | MODELED / INFERRED |
Meta raised 2026 capex guidance to $125B–$145B on Q1 2026 earnings call (April 2026), citing higher component pricing and additional data center costs | Bloomberg reports Meta Compute is in development with two business lines: hosted model access (Bedrock-style) and raw compute capacity (neocloud-style). Reuters separately confirmed the report; Meta declined to comment. | External compute monetization could convert capex from pure expense into revenue optionality — but only if Meta builds the control-plane layer (IAM, billing, compliance, SLAs) that makes cloud services sticky |
Meta confirmed compute partnerships with CoreWeave ($14.2B deal 2025 + $21B deal April 2026), Google, and Oracle in public earnings disclosures | CoreWeave and Nebius sold off sharply on the day of the Meta Compute report — the market read Meta Compute as a substitution risk, not just an infrastructure expansion signal | Contracted neoclouds with anchor customers are validated today but face substitution risk on a multi-year horizon if Meta Compute scales. Uncontracted spot GPU rental markets are the near-term pressure point. |
Meta's MTIA custom inference chip roadmap confirmed: MTIA 300 in production, MTIA 450/500 focused on inference; Broadcom assisting on design, TSMC fabricating (Reuters, March 2026) | Qualcomm confirmed as Meta's first Big Tech customer for data-center CPUs, with deployment expected from late 2028 (Financial Times) | The longer Meta sells compute externally, the more incentive it has to lower cost-per-token via custom silicon — turning what was an internal efficiency project into a margin-defense strategy |
BEAR CASE SPOTLIGHT AWS, Azure, and Google Cloud spent 20 years building the enterprise trust, compliance certifications, developer tooling, and global edge infrastructure that makes cloud sticky. Meta has zero enterprise sales DNA. Its brand is consumer social media — not exactly the reassurance a Fortune 500 CIO needs when putting production workloads on a vendor's infrastructure. Meta's open-source AI strategy creates developer mindshare but fundamental tension with high-margin cloud: if LLaMA is free, Meta has to monetize compute or proprietary models, and Nvidia's cut makes raw compute economics very thin. Meanwhile, $125B–$145B in capex before a dollar of external cloud revenue is an enormous overhang. If the cloud business takes five years to matter, Meta's shareholders will have paid for it twice — once in diluted earnings and once in elevated rates on capital deployed. The Bloomberg report was unconfirmed, Meta declined to comment, and Reuters could not independently verify it. Markets moved on a plan that may change. |
FIVE THINGS TO DO WITH THIS INFORMATION
1. The market is not pricing a social media company. It is pricing an AI infrastructure company with a monetization catalyst it didn't have 48 hours ago. The intraday move is not irrational — it is the market repricing optionality that was previously assigned near-zero value. The capex is not new. The revenue narrative is.
2. Watch the CoreWeave relationship more closely than the Meta headline. CoreWeave is the canary. Meta's $35 billion in committed CoreWeave spend validates the AI cloud model — and makes Meta the customer that could become a competitor. The next CoreWeave earnings call tells you whether Meta is expanding alongside CoreWeave or beginning to displace it.
3. Don't anchor on AWS's moat in the AI-native segment. AWS's enterprise infrastructure moat is real and deep. But the AI developer market operates on different logic: model quality, latency, and price matter more than a 15-year enterprise relationship. Meta's LLaMA distribution is already enormous. A hosted API puts a revenue layer on top of that distribution — in the segment where AWS is most contestable.
4. The custom-silicon stack is the long-game tell. If Meta turns external compute into a margin business, cost-per-token becomes a business model — not just an efficiency target. Watch MTIA deployment pace, Broadcom design-win disclosures, and the Qualcomm data-center CPU timeline. Each is a signal that Meta is building toward owned-margin infrastructure, not just rented-GPU economics.
5. Nvidia wins the buildout. The longer thesis is more nuanced. More cloud infrastructure means more accelerator demand in 2026 and 2027. Full stop. But every dollar of MTIA inference deployed is a dollar that doesn't go to Santa Clara. Nvidia is the near-term winner and the medium-term variable. Price that distinction into the multiple.
The AI Buildout Has a Physical Layer
Many of today’s data centers are still using copper wiring. The same metal we’ve been using for a hundred years.
At the speeds AI demands with data moving between thousands of GPUs, billions of times a second, copper doesn’t just slow down.
It turns that data into heat. The more you push through it, the worse it gets. There’s no software for fix for that.
So what’s the answer?
Explore the Photonics Layer…..
Tuttle Capital Pure Play Photonics ETF (FOTO)
Distributor: Foreside Fund Services | Investing involves risk including possible loss of principle.
News vs. Noise: What’s Moving Markets Today
The news. Warsh’s Fed is not just a hawkish Fed. It is a different operating system. At Sintra, he said the Fed will stick to 2% inflation and “disappoint” anyone expecting tolerance above that level. He also gave almost no forward guidance. That matters more than today’s payroll guessing game. The market wants an answer key. Warsh is taking the answer key away. At the same time, AI semis are giving us the cleanest version of the current market: the fundamental shortage is real, and the trade is crowded. SK Hynix just committed $64.38 billion to NAND and packaging plants, South Korea wants to double memory capacity within five years, and Reuters says AI hyperscaler demand has created a global shortage with NAND and DRAM prices at historical highs. But SK Hynix still fell 15% and Samsung fell 9% after Meta’s reported plan to sell excess AI compute raised capacity fears. That is the whole market in one day. Real theme. Bad entry point. (Reuters — Warsh, Reuters — SK Hynix, Reuters — Meta)
The noise. The noise is the binary argument. Either AI is a bubble, or AI is a supercycle. Wrong. It can be both. Reuters says Torsten Slok told central bankers in Sintra that if AI overdelivers it can hit financial stability, and if it underdelivers it can hit financial stability. WSJ reported the same worry from economists: hyperscaler debt, investor leverage, labor displacement, cybersecurity, and stretched valuations. That does not kill the AI trade. It makes it more selective. Memory, power, storage, networking, and advanced packaging are selling picks and shovels into a capacity race. But the Mag 7, cloud spenders, neoclouds, and “AI attached” momentum names are not all the same trade. The media wants the easy headline. Bubble or boom. The market is telling you the harder truth: bottlenecks can work while the broad AI tape gets dangerous. (Reuters — AI risks, WSJ — economists on AI, MarketWatch — memory selloff)
The dumb advice. The dumb advice is cutting oil forecasts because Hormuz flows “normalized.” WSJ cited OCBC cutting Brent forecasts through 2027 because shipping traffic through the Strait picked up and expectations of normalized flows revived the oversupply narrative. Reuters says oil fell after U.S.-Iran talks in Doha made “positive progress,” but it also says there was no sign of lasting peace, the next meeting is after July 9, and Iran still wants international recognition of control over the strait and the right to levy fees. That is not normal. That is a tollbooth with missiles nearby. This is the same mistake investors made with the Red Sea. A ceasefire can lower the headline risk before it fixes the logistics, insurance, routing, and military risk. Oil can trade lower. Fine. But building a multi-quarter forecast on “normal flows” through a contested chokepoint is not analysis. It is wishful spreadsheeting. (WSJ — oil forecasts, Reuters — oil and Hormuz)
Concrete takeaways.
Warsh is removing the market’s Fed crutch. Less guidance does not mean easier policy.
Memory scarcity is real. Crowded semi positioning is also real.
Buy the bottlenecks, not the whole AI basket.
Treat “AI bubble” and “AI supercycle” as overlapping conditions, not opposites.
Watch hyperscaler debt and excess compute. That is where the ROI question shows up first.
Do not assume Hormuz is fixed because oil stopped panicking.
Long-duration bonds near 5% are not safety. They are a bet Warsh blinks.
Where Does the Money Go When AI Hits a Wall?
When capital chases a tech theme, it tends to pile into the most obvious
layer and miss the one underneath. AI spending is now bumping hard
against memory. Hyperscalers — the big cloud builders like Amazon,
Google, and Microsoft — have shifted memory from 8% of their build
budgets to an estimated 30% in a single cycle. That capital has to go
somewhere. If the constraint is memory, and the build can't move without
it, shouldn't an investor own the layer AI runs on?
View HBMX fund holdings →
Distributor: Foreside Fund Services | Investing involves risk including
possible loss of principal.
<Link = http://www.hbmxetf.com/>
ETF News
A Stock I’m Watching
Hesai is a picks-and-shovels “eyes of robots” play: humanoids, robotaxis, robovans, warehouse/logistics bots, lawn robots, industrial automation, and physical AI all need reliable 3D perception, and Hesai is already positioning itself as a core supplier across those categories. The latest numbers support the theme: in Q1 2026, Hesai’s robotics lidar shipments rose 137.8% year over year to 118,282 units, total lidar shipments rose 140.9%, revenue grew 29.6%, and the company stayed GAAP profitable, which helps separate it from many speculative robotics names that still lack scale. The stock also has fresh narrative fuel: Hesai says its lidar is powering Honor’s humanoid robot “Lightning,” it secured an exclusive 200,000-unit design win with Zelos, is Neolix’s largest lidar supplier, and is pushing new “spatial intelligence” products like Kosmo for robotics simulation and training. The risk is that HSAI still carries China/ADR and U.S. geopolitical overhang, including its ongoing dispute around the U.S. Defense Department’s Chinese military-company designation, but that overhang is also part of why the name can move hard when sentiment improves. If the market is rotating back into robotics and physical AI, HSAI gives direct exposure to the sensor layer that many robots will need, with real shipments, profitability, and customer wins already backing the story.
In Case You Missed It
Great conversation on wide ranging topics with Kenny Polcari…
The H.E.A.T. (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.