Missed NVDA? AMD's 107% and Marvell's 46% Data-Center Jumps Mark the Next Earnings Inflection

September is living up to its seasonal reputation; the weekly dashboard shows S&P 500 down ~8% on the week yet +12% year-to-date, Nasdaq +7% on the week and +14% YTD as the only positive index, Dow −1.7% on the week and +7.5% YTD lagging, Russell 2000 −1.5% on the week and near −5% in two weeks yet +15% YTD . The video's warning sits above those numbers: September seasonality plus a fresh Fed move keeps volatility elevated. The headline is the move itself — Fed funds target up 25 bps to 3.75–4%, first hike since July 2023, 12–0 unanimous — with the statement noting inflation remains elevated and a timelier return to 2% as the intent, while also saying economic activity expands at a solid pace, domestic spending resilient, productivity strong, capital spending robust . The swing from a year that began pricing cuts to now pricing a hike is a direct discount-rate headwind for long-duration growth names.

Rates, Bonds and Growth: Why the Discount Gets Heavier

When rates rise, money costs more — firms borrow dearer, consumers finance dearer, and investors discount future earnings at a higher rate. A growth stock priced on profits five or ten years out loses present value fast when that discount rate climbs; add rated bonds offering 7% coupons (not junk) as a competing yield and equities must compete with bonds again. The video flags that 7% on investment-grade is not hypothetical — it is on screens — so the equity risk premium reprices quickly. The lesson is price: even a durable long-term thesis needs an entry that respects the new cost of capital, a point the video returns to when sizing AMD and Marvell.

Housing is where tightness is most visible. 30-year mortgages above 7% meet historically elevated prices for a brutal affordability equation ; existing-home sales around 4 million annualized mark one of the weakest volumes since 2011 . Prices are not collapsing, transactions are — a transaction recession . Millions who locked 2–3% (the host cites 2.6%) will not swap into 7% , constraining supply ; new buyers cannot carry the payment, constraining demand . Reuters' analyst survey sees mortgages averaging ~6.6% over two quarters with only modest price growth. For the Fed the trade is awkward: housing struggles while the broader economy, in its own words, keeps expanding — a split that keeps market swings alive.

What the Businesses Are Actually Saying

Strip the headlines and ask the businesses. Nvidia Q2 FY27: $96.2B revenue, up 106% YoY, Data Center $89.0B up 117% — both gross margins 75.0% , diluted EPS $2.46 GAAP / $2.22 non-GAAP ; Broadcom reports similarly strong AI semiconductor demand and hyperscalers keep building. Instead of “is AI over?” the better question is “where does the next earnings growth come from?” Goldman Sachs expects global AI investment above $1 trillion in 2026 ; Value Add Pulse tallies ~$725B capex guided by Alphabet, Amazon, Meta and Microsoft in 2026, up 77% from $410B , with Amazon ~$200B, Microsoft ~$190B, Alphabet $175–185B, Meta $115–135B . The opportunity is no longer buy a GPU , it is compute + custom silicon + networking + memory + power — an ecosystem , not a single chip. Think of it as moving from pouring the foundation to fitting the building.

AMD: +50% Total, +107% Data Center

AMD sits where that fitting happens. Q2 2026 total $11.5B (+50% YoY), Data Center $6.7B (+107%) — ~58% of revenue , driven by EPYC and Instinct (Converge Digest / RuntimeWire). The more telling line is operating leverage : Data Center operating income $2.1B, margin 31% , versus an operating loss a year ago (comparison affected by an $800M export-control charge ). That shift — revenue up, margins up faster — is the core of the Nvidia-2023 analogy . Then, Nvidia FQ1 FY24 $7.2B total / $4.3B Data Center, guided ~$11B next quarter , and delivered $13.5B total (+101%) and $10.3B Data Center (+171%) , a print that broke the “normal cycle” narrative. Today AMD mirrors the setup: growth accelerating, estimates moving up, hyperscaler deployments expanding .

The answer is a platform, not a chip . Helios (rack-scale AI — GPUs + CPUs + networking + software in one rack) bundles Instinct MI450 GPUs with CPUs and software. Customers are not theoretical: AMD-Anthropic strategic partnership for up to 2 gigawatts of MI450 in Helios racks (MD and IR releases, July 22, 2026) and an expanded Microsoft–AMD Helios + 6th-gen EPYC rollout across Azure . Add Amazon, Meta, Oracle, Alphabet as overlapping customers and even a modest share of a hundreds-of-billions market matters. The video shows an Edge score 71 and price targets $600–710 ( ~40% upside at $710, author sees $1,000+) ; it also discloses an Oct 16 $370 put sold for $350, now $60 to buy back — $290 (≈83%) per contract — and an Options Edge Plus 85% hit rate (100% in August) as the options overlay. Ownership plus overlay is the tell that this is a position, not a slide.

What AMD must still solve is the CUDA wall (Nvidia's software moat — the language most models and libraries run on) . Hardware parity alone does not move workloads; software migration, networking scale, and ecosystem take time. The right lesson from Nvidia is not “buy anything growing.” It is a checklist: revenue accelerating, estimates rising, market expanding, products deployed, margins widening, competitive position strengthening . For AMD that means Instinct/MI450 ramp, Helios at scale, CPU share gains, narrowing the software gap, and flow-through to operating income — earnings power years out could look very different from today's print if those line up.

Marvell: The Custom-Silicon Architect

Marvell (MRVL) slices the same pie differently: custom silicon and networking . If Nvidia builds a Ferrari (general-purpose accelerator) and AMD builds a McLaren , Marvell helps the biggest tech firms build their own cars . When you spend tens of billions a year , a workload-specific chip can improve performance, power efficiency, cost and control — but designing at the leading edge is exceptionally hard, with Marvell and Broadcom as the few who can. The quarter reads well: record $2.739B revenue (+37–40% YoY), Data Center +46%, GAAP gross margin 53.1% / non-GAAP 58.9%, GAAP EPS $0.33 / non-GAAP $0.94, operating cash flow >$600M , management attributing strength to the data center portfolio . That is AI infrastructure pulling the business forward , not a one-off.

The turn is the Alphabet-Google expansion . The companies signed a commercial agreement covering the TPU ecosystem — inference chips, storage and network controllers, memory interfaces and near-memory compute blocks — with warrant incentives for Marvell. SEC 8-K and CNBC (Aug 19, 2026) detail up to $12.2B in shares available to Google and a ~6% pop on the news ; this is not a single component but multiple layers of Google's custom stack . “Is spending slowing?” The video answers with behavior, not slogans: a customer issuing multi-billion-dollar warrants this quarter is not slowing. The investor page flags sharp multi-year earnings growth if near consensus holds and latest targets $270–325 ( ~35% upside at $325, author sees year-end) , the price mirror of that commercial step-up.

Same Wind, Two Sails

Place them side by side to reduce error. AMD is the direct Nvidia-adjacent bet — accelerators + CPUs + racks + software + networking , winning on share and scale . Marvell is the pick-and-shovels bet — custom silicon, connectivity, networking, storage and memory interfaces behind hyperscaler-specific architectures . AMD risk = product ramp + software ; Marvell risk = customer concentration + custom-silicon margin math . The video's stance — own both, add via options — frames them not as rivals but as two sails for the same wind , the AI infrastructure buildout .

The Nvidia Lesson: Track Trajectory, Not Hype

Nvidia's lesson is trajectory, not hype. From ~$7.2B quarterly revenue pre-inflection (FQ1 FY24) to ~$100B in a recent quarter , the company traced what a once-in-a-generation architecture shift can do when Wall Street underestimates the market . The video's caution is explicit: for every Nvidia there are dozens where extrapolated growth never arrived . Watch earnings trajectory . If AMD captures meaningful accelerator share and Helios ships at volume , or Marvell's custom programs scale to tens of billions , the market will price it — and early means before the street fully does . The macro footnote stays: Fed 3.75–4%, mortgages >7%, 4M transactions remain a valuation headwind ; the cycle can be huge and entry still matters.