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 $4 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.
Nvidia’s $150M “Tell”: The AI Gold Rush Is Moving From Training to Inference
The story Wall Street is missing
For the last two years, “AI investing” has basically meant one thing: training. Build bigger models. Buy more GPUs. Pour more concrete. Burn more power.
That trade worked because training is the obvious bottleneck when everyone is racing to build the smartest brain.
But here’s the regime shift: training makes headlines… inference makes money.
Inference is the unsexy part of AI: the moment your model is actually used in production—answering customer questions, routing deliveries, writing code, detecting fraud, optimizing ads, and running workflows all day long. It’s where AI stops being a demo and becomes a utility bill.
And Nvidia just tipped its hand.
This week, Nvidia reportedly put $150M into Baseten as part of a $300M round at a ~$5B valuation—Baseten’s business is helping companies deploy and run AI models in real-world production environments.
That’s not a random venture dart throw. That’s Nvidia saying: “The next fight is where AI runs… not where AI is trained.”
Baseten literally markets itself around this idea—“Inference is everything.”
Why this matters: training is CapEx… inference is a perpetual toll
Training is episodic. You train a model, upgrade it, train again.
Inference is constant. If AI becomes embedded in every workflow, inference becomes a recurring consumption business—measured in tokens, calls, latency, uptime, cost per answer, and watts per output.
That shift changes everything:
The profit pool moves up the stack (deployment, routing, optimization, orchestration).
The competitive set widens (GPUs vs TPUs vs ASICs vs “good-enough” accelerators).
Efficiency becomes the battlefield (cost per token, tokens per watt, latency per dollar).
In other words: the market is transitioning from “Who has the biggest brain?” to “Who can run the brain the cheapest… at scale… without breaking the grid or the budget?”
Nvidia’s real message: “We’re not just selling chips anymore”
If inference becomes the dominant workload, it creates a dangerous opening:
Training has been Nvidia’s fortress. Inference is where challengers can say:
“You don’t need a Ferrari to drive to the grocery store.”
That’s why Nvidia is building a full-stack moat—not just GPUs, but the software, tooling, and now the deployment layer where enterprises decide what runs where.
It also explains the other chess move: Nvidia’s reported licensing/talent deal around Groq, an inference-focused chip company—one report pegged it as a massive transaction (around $20B) tied to inference IP and key hires.
Whether the exact number ends up being right matters less than the direction: Nvidia is spending aggressively to keep inference from becoming the crack in the dam.
This is the playbook of every platform monopoly in history:
When the next layer becomes strategic, you don’t compete there… you own it.
What it foreshadows for AI in 2026
1) “AI adoption” becomes measurable (and painful).
2026 is when boards stop asking, “Do we have an AI strategy?” and start asking:
“What is our ROI?”
“What did inference cost us this quarter?”
“Did margins improve or did we just buy fancy autocomplete?”
2) Inference turns into the new bottleneck.
When usage explodes, the problem stops being “Can we train it?” and becomes:
“Can we serve it fast enough?”
“Can we serve it cheap enough?”
“Can we serve it reliably enough?”
3) AI becomes a cost war.
If everyone has access to similar model capability, differentiation shifts to:
orchestration,
optimization,
proprietary data/workflows,
distribution,
and unit economics.
That’s why inference specialists (like Baseten) suddenly matter.
Winners: who gets paid in an inference-first market
Core winners (direct beneficiaries)
NVDA — Not just “GPU king.” This is Nvidia explicitly defending the inference profit pool and trying to own the deployment layer.
AMD — Inference is where buyers get price-sensitive and alternatives get a real shot. AMD doesn’t need to “kill Nvidia” to win; it just needs to take share at the margin in a bigger market.
AVGO — Custom silicon is the quiet inference monster. If inference becomes the dominant workload, ASICs and custom accelerators get a larger seat at the table.
ANET — Inference at scale is a networking story. More distributed serving = more east-west traffic = more demand for high-performance data center networking.
MU — If inference is everything, memory bandwidth and supply remain a gating factor. Memory pricing power becomes a feature, not a bug.
VRT — Inference scaling isn’t just chips—it’s power delivery, cooling, and keeping racks alive at higher utilization.
Second-order winners (the “toll booths” around inference)
NET / AKAM — If inference pushes toward the edge (latency, cost, distribution), edge platforms gain strategic relevance.
EQIX / DLR — More inference nodes, more colocated compute, more interconnection demand.
Losers: who gets squeezed as inference becomes ubiquitous
1) “AI veneer” software (seat-based pricing with weak workflow lock-in)
If AI agents can replicate features cheaply, the market will punish software that can’t prove it owns a workflow end-to-end. The near-term tape risk tends to show up first in big incumbents even if they ultimately adapt:
CRM, NOW, ADBE, INTU — Not because they’re “dead,” but because they’re forced to defend pricing and prove monetization in a world where AI assistants keep getting more capable.
2) Consumer electronics that eat rising component costs
If memory and compute keep getting bid up by data centers, someone else pays:
AAPL, HPQ, SONY — margin pressure risk if component inflation rises faster than consumer pricing power.
3) “Middlemen” infrastructure with no moat
As inference tooling improves, any provider whose only pitch is “we rent GPUs too” faces margin compression. (This is less about one ticker and more about a business model problem.)
The actionable takeaway for investors
If you remember one thing, make it this:
Training was the land grab. Inference is the tax base.
So don’t only own the “AI builders.” Own the businesses that get paid when AI is used millions of times per day—and the companies that make AI cheaper, faster, and more deployable.
And keep a hedge mindset: when the market transitions from build to run, leadership changes fast—and dispersion gets violent.
News vs. Noise: What’s Moving Markets Today
If gold is ripping while yields are rising, that’s not classic risk-off.
Classic risk-off is: yields down, dollar up, stocks down.
What you’re describing is: policy/credibility risk — where investors want hard-asset insurance (gold), while simultaneously demanding more compensation for holding long-term paper (higher yields / higher term premium). That’s the signature of “the shock is coming from Washington (or policy), not from growth fundamentals.”
What it means for positioning (the concrete takeaways)
This is how I’d frame it — simple rules, not predictions:
Stop watching Greenland. Start watching the long bond.
The market’s “truth serum” is the long end. If the 10s/30s keep backing up while the dollar can’t catch a bid, that’s a warning that global investors are charging the U.S. a credibility premium.Japan is the match near the gasoline.
A disorderly move in super-long JGBs doesn’t stay in Japan. It leaks into USTs via global asset allocation, hedging flows, and the “who funds the deficits?” question. If Japan stays hot, U.S. duration stays vulnerable even if the S&P bounces.Affordability politics = bond-market fragility.
If the administration needs lower rates to support housing/consumer optics into midterms, it cannot afford a sustained bond tantrum. That makes the “threat → wobble → walk-back” pattern more likely as long as the bond market keeps biting. In other words: the bond market is Trump’s constraint.The “repetition ≠ fact” lesson is spreading (and it matters).
The Intel example is the broader 2026 disease: momentum + narratives + passive flows can levitate a stock… until the business has to perform. The market is starting to re-learn this lesson across pockets of AI capex, semis, and “story stocks.” When yields rise, that re-learning accelerates — because high rates are gravity.Hedges aren’t optional in this regime — they’re admission tickets.
You don’t hedge because you’re bearish. You hedge because headlines now create policy-volatility air pockets, and the bond market can turn those air pockets into real damage.Gold remains the cleanest “U.S.-policy shock” hedge.
Rate/Duration awareness matters more than ever (watch auctions, term premium behavior, curve moves).
If you’re heavy long-duration growth, consider defined-risk hedges (spreads/ratios) rather than naked fear trades.
Who wins / who loses if this is truly a “bond vigilante” tape
Likely winners (if yields stay jumpy):
Hard assets / precious metals (insurance bid when credibility is questioned)
Quality cash-flow cyclicals (less multiple-risk than long-duration growth)
Financials (selectively) if higher yields = better NIM and credit doesn’t crack
Energy/materials/industrials tied to the “real economy” rotation you’re already seeing
Likely losers:
Long-duration “priced for perfection” growth (multiple compression risk)
Heavily levered consumers (financing costs + savings-rate pressure)
Crowded AI capex proxies if the market starts asking “what’s the ROI?” while discount rates rise
A Stock I’m Watching
Today’s stock is ASML…..
ASML is the “stock I’m watching” in semicap because the demand signals are getting increasingly lithography-specific: Wells Fargo recently noted that China’s December semiconductor tool imports were up +83% m/m, driven by record lithography, even while 4Q25 tool imports were down -13% q/q (-6% y/y)—a setup that screams “chokepoint + pull-forward” rather than broad-based weakness. In that context, ASML remains the cleanest way to express “leading-edge keeps compounding” (TSM/Samsung/Intel node ramps ultimately bottleneck at litho), but it also carries the most headline/timing sensitivity (export controls, customer pull-ins, and inevitable digestion quarters). Wells Fargo has ASML Overweight with a $1,450 PT and keeps it high on their semicap preference list (AMAT, LRCX, ASML, KLAC). What I’m focused on next: whether this “record litho” surge is a one-off pre-buy or a sign the litho layer is tightening again, and how ASML frames the mix (China vs. non-China), plus order/backlog tone that can validate the next leg of the WFE upcycle.
In Case You Missed It
Markets don’t just move on fundamentals—they move on themes, positioning, and policy. In this episode, I’m joined by Matt Tuttle, founder of Tuttle Capital Management, to break down his “HEAT” framework: Hedge, Edge, Asymmetry, and Themes—and how those four ideas shape his daily approach to risk and opportunity. We discuss why Matt believes you should always be hedged, what real “edges” look like (and why many disappear once Wall Street markets them), how to structure trades for asymmetric payoffs, and how he’s thinking about 2026. Topics include the shift from AI creators to AI adopters, the importance of AI capex and the Fed as key pillars supporting the market, and how policy shocks can create both landmines and upside.
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.© 2025 Tuttle Capital Management, LLC (TCM). TCM is a SEC-Registered Investment Adviser. All rights reserved.