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.
Now Trading: Tuttle Capital Magnificent 7 Income Blast ETF
$MAGO combines direct exposure to the Magnificent 7 with a systematic put‑spread program implemented with listed options and FLEX Options®.
Distributor: Foreside Fund Services
Learn More: IncomeBlastETFs.com
I’m hosting a webinar entitled “Why Covered Call ETFs Suck and What to Do Instead” (More Info Below) January 15 2-3pm. Sign Up Here
Table of Contents
H.E.A.T.
NVDA announced a major deal last week. This morning I talk about why, and the implications for the AI trade……
Nvidia–Groq Deep Dive: What This Deal Really Means for 2026
The Setup: “Why would Nvidia need anyone else’s chip tech?”
Because the AI market is quietly changing shape.
Training made Nvidia a legend.
But inference (the everyday “answer my query / run my agent / generate my response” workload) is where volume explodes — and where latency + memory behavior start to matter as much as raw compute.
And that’s why this Groq deal is important: it’s Nvidia planting a flag in the part of the AI stack that was supposed to be where hyperscalers and ASIC startups finally chipped away at the GPU monopoly.
What actually happened (and what didn’t)
Here’s what we can say cleanly from the public record:
Nvidia and Groq signed a “non-exclusive inference technology licensing agreement.”
As part of that agreement, Groq’s founder/CEO Jonathan Ross and certain “key employees” are joining Nvidia (Groq emphasizes this is to help integrate the licensed tech).
GroqCloud continues operating independently (Groq explicitly states its inference cloud remains separate).
And now the messy part — the rumor mill:
Some reporting framed this as a ~$20B “acqui-hire” / purchase of assets, but Nvidia indicated it was not an acquisition and didn’t confirm a price.
Bottom line: This is not Nvidia buying Groq outright (at least as described publicly). It’s a license + talent migration + integration play — the modern Big Tech “deal without calling it a deal.”
The real story: Groq is an inference weapon, not a training rival
Groq’s core pitch has always been:
“Stop treating inference like training.”
GPUs are incredible generalists. But inference has different pain points:
Training:
Big batches
Throughput matters
You can tolerate more latency
You care about total compute delivered
Inference (especially where the world is going):
Interactive + user-facing
Latency is product quality
Memory bandwidth + memory locality become gating factors
The bottleneck shifts from “how fast can you compute” to “how fast can you feed the compute and return tokens”
This is why your note about SRAM-heavy LPUs is the key lens:
Groq’s architecture leans into embedded memory (SRAM) and deterministic execution to drive ultra-low latency decoding — the part of inference where the model is generating token after token and users are literally waiting.
If the next wave of AI products is agents, voice, real-time copilots, and speculative decoding / chain-of-thought-style reasoning, then milliseconds matter.
That’s the bet.
Why Nvidia did this now: 3 strategic motives
1) Inference is where Nvidia could actually lose share
Hyperscalers already have (or are building) inference silicon. Startups are targeting inference because it’s the one place you can credibly say, “GPUs aren’t optimal.”
A non-exclusive license + integration lets Nvidia say:
“Fine. If inference becomes ASIC-land… we’ll be the platform that sells ASIC-like inference too.”
2) This is a “defense that turns into offense”
This is the Instagram playbook:
You don’t wait until the upstart becomes huge.
You neutralize it early, fold it into your ecosystem, and make it complementary.
Even if Groq was never going to dethrone Nvidia, it represented a real narrative threat:
“Inference doesn’t need GPUs.”
Nvidia cannot allow that meme to compound.
The future data center isn’t “one chip to rule them all.”
It’s a rack-level architecture problem:
Different workloads want different cost curves:
tokens/$
tokens/watt
latency per token
throughput per rack
memory bandwidth constraints
A world where Nvidia sells a hybrid stack (GPU for training + specialized inference engines for decode-heavy workloads) is a world where Nvidia expands TAM and keeps customers inside the Nvidia “store.”
The 2026 implication nobody wants to say out loud: margins
If this deal works, it’s bullish for Nvidia’s strategic control…
…but it also hints at a reality shift:
The more compute becomes commoditized
The more inference shifts to specialized silicon
The more pricing pressure shows up
…the harder it is to maintain “golden-era” GPU-like margins forever.
This doesn’t mean Nvidia “falls.” It means:
The next inning is about mix, attach, and platform economics — not just GPU scarcity.
Winners: Who benefits if Nvidia + Groq becomes “hybrid inference at scale”
Tier 1 winner: $NVDA
Owns the distribution.
Owns the software gravity.
Now adding a latency/inference arrow to the quiver.
Tier 2: The “inference accelerant” ecosystem (pick-and-shovels)
These benefit if inference volumes explode and racks become more heterogeneous:
Networking / interconnect:
$AVGO (switching / connectivity exposure)
$ANET (data center networking plumbing)
Data-center power & thermal (because inference scaling is still watts + heat):
$VRT (critical power + cooling)
$ETN (power distribution and electrical gear)
Tier 3: Foundry / manufacturing angles (watchlist logic)
If Groq-style silicon ramps and Nvidia starts offering “more than GPU” configurations, you get leverage to whichever foundry + packaging chain gets the work.
Your note mentions Groq’s earlier gen at $GFS and potential future work at Samsung. Treat this as watchlist, not “guarantee.”
Losers: Where this hurts (or at least raises the bar)
1) Inference-only startups (public or future IPO bait)
If your entire pitch was:
“We do inference better than Nvidia”
…Nvidia just told the market it will partner/license/hire its way into your lane.
This doesn’t kill the category — but it compresses the odds of a clean standalone win.
2) GPU rental middlemen with weak differentiation
If inference starts moving to specialized silicon, then “rent GPUs to run inference workloads” can become a worse business:
Lower utilization quality
Price competition
Customer migration to cheaper inference stacks
(Translation: the “CoreWeave-style” logic becomes riskier when the workload mix changes.)
3) $INTC (as a narrative loser right now)
Not because Groq is “Intel vs Nvidia,” but because this reinforces the market’s belief that Nvidia is the control plane — and everyone else is trying to catch up on cadence, ecosystem, and integration.
Also: your note about Nvidia no longer testing 18A would be another narrative headwind — but I’m treating that as “reported / watch for confirmation,” not something to anchor the thesis on.
This is the first big clue that AI is splintering
In 2025 you bought “AI exposure.”
In 2026 you’re buying one of three businesses:
Manufacturers (sell the compute)
Spenders (buy the compute)
Monetizers (turn AI into revenue and margin)
This Groq deal is Nvidia saying:
“We intend to stay the manufacturer even as inference becomes its own battlefield.”
What I’m watching next (simple checklist)
Does Nvidia productize this?
Licensing is cute. Shipping integrated systems is the real tell.Does Groq tech become a “line item” inside Nvidia’s platform story?
If Jensen starts talking about inference latency the way he used to talk about CUDA, that’s the signal.Any sign of margin narrative shifting?
Not “down,” but “mix pressure.” Watch how the street starts modeling it.Does this trigger more “soft acquisitions”?
If yes, it confirms the arms race is moving from chips → systems → vertical integration by necessity.
The Nvidia–Groq deal isn’t about Nvidia being scared.
It’s about Nvidia being realistic:
Training made the kingdom.
Inference decides who collects taxes.
And Nvidia just hired one of the smartest “inference rebels” on the board to make sure the next AI boom doesn’t happen outside the Nvidia ecosystem.
News vs. Noise: What’s Moving Markets Today
The noise yesterday was the continued parabolic moves in precious metals. I say noise because while I think you need to be long precious metals, I think you probably avoid this volatility and don’t buy dips or rips. Once things settle down, meaning nobody is talking about gold or silver anymore, then you probably go in. I mentioned yesterday that I cut my silver position way down on Friday. Yesterday was a tempting dip, but I think you wait a bit longer here.
Meanwhile, the news continues to be interest rates and capex. As long as the Fed is dovish and companies are spending hand over fist on AI then you need to be long this market.
A Stock I’m Watching
Today’s stock is Reddit (RDDT)…..
Reddit (RDDT) is worth watching because it sits at the intersection of two underappreciated secular trends: (1) performance advertising migrating toward “high-intent” surfaces, and (2) AI making authentic human-generated context more valuable, not less. Reddit isn’t just another social feed—it’s an intent graph organized around communities, where users self-identify problems (“best VPN?”, “replace my water heater?”, “side effects of X?”) in a way that’s unusually monetizable when the ad stack and measurement are done right. The upside narrative is that Reddit can compound ARPU by improving ad tooling (targeting, measurement, creative formats, and conversion optimization) while also layering in high-margin, recurring revenue streams tied to data access/licensing as AI developers and “answer engines” increasingly depend on fresh, real-world language and community knowledge. The asymmetry is that the market often values Reddit like a cyclical ad business, but the long-term outcome can look more like a durable “demand capture + data tollbooth” platform if execution continues. The key things to track are ad load discipline (don’t break UX), proof that performance ads scale without degrading community health, and whether data monetization becomes meaningful and repeatable rather than one-off.
In Case You Missed It
How Else I Can Help You Beat Wall Street at Its Own Game
Inside H.E.A.T. is our monthly webinar series, sign up for this month’s webinar below….
Why Covered Call ETFs Suck-And What To Do Instead
Thursday January 15, 2-3PM EST |
Covered call ETFs are everywhere — and everyone thinks they’ve found a “safe” way to collect yield in a sideways market. |
The truth? |
They cap your upside, mislead investors with “yield” that’s really your own money coming back, and often trail just owning the stock by a mile. |
Join me for a brutally honest breakdown of how these funds actually work — and what you should be doing instead. |
What You’ll Learn:
🔥 Why “high yield” covered call ETFs are often just returning your own capital |
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.