Jensen Huang’s $1 Trillion Bet: 5 Stocks That Benefit Most From Nvidia’s GTC 2026 Bombshell

AI Investing & Semiconductors

At GTC 2026, Nvidia’s CEO doubled his demand forecast to $1 trillion through 2027. He then said even that won’t be enough. Here is what it means for investors — and which five stocks are best placed to benefit.

Technology prices normally fall over time. Competition and efficiency push costs down. But AI chips are breaking that rule. And Jensen Huang just confirmed it won’t change soon.

At Nvidia’s GTC conference in San Jose on March 16, Huang made a bold prediction. He now expects at least $1 trillion in purchase orders for Nvidia’s Blackwell and Vera Rubin chips through the end of 2027. One year ago, he said $500 billion. He doubled it — and then said even $1 trillion might not be enough.

Nvidia demand forecast through 2027
$1T+
Up from $500B at GTC 2025 — doubled in one year
FY2026 Revenue
$215.9B
+65% year-on-year
Q1 2026 Guidance
$78B
Beat consensus by $5B+

What Huang Said — and Why It Matters

Huang’s $1 trillion figure is not a guess. It refers to real purchase orders already in the pipeline. These come from major cloud companies, AI labs, and enterprises around the world.

He was direct about the supply gap. “In fact, we are going to be short,” he said. “I am certain computing demand will be much higher than that.”

That tells investors something important. The problem is not demand — demand is already there. The problem is making enough chips fast enough to meet it. That gap benefits every company in the supply chain.

“I see through 2027 at least $1 trillion. In fact, we are going to be short. I am certain computing demand will be much higher than that.”

— Jensen Huang, Nvidia CEO, GTC 2026, March 16, 2026

Nvidia posted $215.9 billion in revenue for fiscal 2026 — up 65% from the year before. It has grown revenue above 55% for eleven straight quarters. The $1 trillion figure is the next chapter of a story that has already proven itself.

What Nvidia Announced at GTC

The $1 trillion forecast was the headline. But there was much more.

Key GTC 2026 Announcements
  • Vera Rubin platform — Nvidia’s next AI chip system. Built on TSMC’s 3nm process. Cuts inference cost per token by 10x vs Blackwell.
  • Vera Rubin NVL72 — The flagship rack. 72 GPUs. 22 TB/s memory bandwidth. 3.3x more inference performance than Blackwell Ultra.
  • Kyber rack architecture — The generation after Rubin. 144 GPUs in a vertical tray. Ships in 2027 as Vera Rubin Ultra.
  • Groq 3 LPU — Nvidia’s first chip from its $20B Groq acquisition. Works alongside Vera Rubin to speed up inference.
  • Autonomous vehicles — BYD, Geely, Nissan, Isuzu, and Hyundai are building Level 4 self-driving cars on Nvidia’s Drive Hyperion platform.
  • Uber partnership — Nvidia-powered robotaxis launch in LA and San Francisco in 2027. Expanding to 28 cities by 2028.

The central theme at GTC was agentic AI. This is AI that does not just answer questions. It plans, reasons, and takes actions on its own. Huang believes this shift will drive far more computing demand than simple chatbots ever did.

He even predicted that software companies will become “AaaS” — agentic AI as a service — selling AI agents instead of software tools. Every agent generates many more computing tokens than a simple chat reply. More tokens means more GPUs. More GPUs means more revenue for everyone in the supply chain.


5 Stocks That Benefit Most

Nvidia is the obvious winner of its own $1 trillion forecast. But smart investors look one step further. They ask: who else captures revenue from this buildout? These five stocks sit directly in the path of that trillion dollars. They benefit whether or not Nvidia keeps every point of its market share.
1
NYSE: TSM
Taiwan Semiconductor (TSMC)
Foundry

The Simple Case

Every Nvidia chip must be made somewhere. That somewhere is TSMC. The Vera Rubin GPU uses TSMC’s 3nm process. The next generation — Feynman — is planned for TSMC’s 1.6nm process. TSMC is not just a supplier. It is the physical bottleneck through which every chip in that $1 trillion forecast must pass.

Why It Wins Beyond Nvidia

Nvidia makes up about 19% of TSMC’s total revenue. But the case is bigger. Google’s TPU chips, Amazon’s Trainium, and Microsoft’s Maia chips are all made by TSMC. When AI companies build their own chips to reduce Nvidia reliance, those chips also go to TSMC. Buying TSMC is like buying the whole AI chip industry in one stock.

The Numbers

TSMC plans to spend $52–56 billion on new factories in 2026 — nearly a third more than last year. About 70–80% of that goes to advanced chip processes. Analysts expect TSMC earnings to grow 36% in 2026. If Nvidia’s orders run hotter than expected, those estimates could prove too low.

Why it benefits from Huang’s $1T forecast Every Blackwell and Vera Rubin chip flows through TSMC’s factories. A bigger Nvidia order book means a bigger TSMC order book. And if hyperscalers build custom chips to reduce Nvidia reliance, those chips go to TSMC too.
2026 EPS estimate
$14.54
Earnings growth
+36%
2026 CapEx plan
$52–56B
Nvidia share of rev.
~19%
2
NASDAQ: MRVL
Marvell Technology
Networking

Nvidia Just Bet $2 Billion on This Company

On March 31, Nvidia completed a $2 billion equity investment in Marvell. The two companies are building NVLink Fusion together. This integrates Marvell’s optical networking directly into the Vera Rubin chip system. It is a strategic partnership — not just a supply deal.

Why the Deal Matters

Marvell also supplies custom AI chips to Amazon, Alphabet, and Microsoft. It sits at the crossroads of Nvidia’s hardware and the hyperscalers’ own chip programmes. When AI companies build their own processors, Marvell’s networking connects them all. That is a very powerful position.

What the Market Said

Marvell’s stock jumped 13% the day the investment was announced. That is a big move for a partnership announcement. The market is re-rating Marvell as a core part of AI infrastructure — not just another chip supplier.

Why it benefits from Huang’s $1T forecast More Nvidia chips shipped means more NVLink Fusion networking needed. Marvell earns revenue from every rack where its optical connectivity runs — across both Nvidia systems and hyperscaler custom chips.
Stock reaction
+13%
Nvidia stake
$2B
Key product
NVLink Fusion
Hyperscaler clients
AWS, Google, MSFT
3
NASDAQ: MSFT
Microsoft
Cloud & AI

The Biggest Buyer and the Biggest Seller

Microsoft plays two roles in the AI boom. It is one of the largest buyers of Nvidia chips. And it is one of the best-placed companies to sell AI services to businesses.

At GTC 2026, Nvidia confirmed that Microsoft’s new Fairwater AI superfactories will run on Vera Rubin NVL72 rack systems. Microsoft has said about two-thirds of its total capital spending goes to GPUs and CPUs. With Nvidia holding 85–90% of the AI GPU market, most of that money goes straight to Nvidia’s products.

How Microsoft Earns from AI

Microsoft is not just buying chips — it is selling AI. Through Azure cloud, Copilot tools, and enterprise software. Every business that runs an AI agent runs it on Azure. Every token that agent produces earns Microsoft cloud revenue. As agents handle more tasks, Azure usage grows.

The Safety Net

Microsoft also has Office, Windows, and Xbox. These protect investors if AI spending slows. No pure-play chip stock can offer that buffer. For investors who want AI upside with less downside risk, Microsoft is the most balanced choice on this list.

Why it benefits from Huang’s $1T forecast Microsoft buys Nvidia chips at scale and earns cloud revenue from every AI workload on Azure. The agentic AI shift Huang described at GTC is a direct boost to Azure billing. More agents mean more compute hours sold.
CapEx on GPU/CPU
~67%
AI platform
Azure + Copilot
GTC announcement
Fairwater superfactories
Revenue streams
Cloud + Software + AI
4
KRX: 000660 / OTC: HXSCL
SK Hynix
HBM Memory

The Memory Bottleneck

The Vera Rubin GPU needs a new type of memory called HBM4. It is the most advanced memory ever put in a commercial chip. It delivers 22 terabytes per second of bandwidth per GPU. The Rubin rack carries 2.5 times more memory than its Blackwell predecessor.

Right now, only one company can supply HBM4 in volume: SK Hynix.

Why This Position Is Strong

SK Hynix has supplied HBM memory to Nvidia since the H100. It moved fastest to develop HBM4. As each new Nvidia chip needs more and faster memory, SK Hynix’s leverage grows. Huang’s $1 trillion chip demand forecast is also, in practice, a huge demand forecast for HBM4.

U.S. Listed Option

SK Hynix trades on the Korean Stock Exchange and via OTC markets. For a more liquid alternative, Micron Technology (MU) is the U.S.-listed option. Micron is qualifying its own HBM memory for Nvidia. Either stock gives you exposure to the memory shortage that Huang named as a key supply constraint.

Why it benefits from Huang’s $1T forecast Vera Rubin needs 2.5x more DRAM and 1.5x more HBM per rack than Blackwell. A bigger chip pipeline means a much bigger memory pipeline. SK Hynix and Micron are the only companies that can fill it.
Memory supplied
HBM4
Rubin DRAM uplift
2.5× vs Blackwell
Rubin HBM uplift
1.5× vs Blackwell
US listed alt.
Micron (MU)
5
NASDAQ: CRWV
CoreWeave
AI Cloud

Named at GTC — First in Line for Rubin

Nvidia named CoreWeave directly at GTC 2026. CoreWeave will be among the first cloud providers to offer Rubin-based systems — starting in the second half of 2026. This is not by chance. Nvidia holds a significant equity stake in CoreWeave. The two companies are closely linked.

What CoreWeave Does

CoreWeave rents GPU computing power to AI labs, startups, and enterprises. This is the same business model the H100 price surge confirmed is working. H100 rental prices jumped nearly 40% in six months. Spot capacity sold out completely. CoreWeave’s locked-in GPU inventory has become one of the most valuable assets in tech.

The Risk to Know

CoreWeave recently completed its IPO. It is scaling fast and spending heavily. Execution risk is real — this is not a slow, steady business. But for investors who can handle that risk, the upside potential is significant. No other listed stock gives this direct an exposure to the AI compute rental market.

Why it benefits from Huang’s $1T forecast More demand for AI compute means more companies renting GPUs instead of buying them. CoreWeave holds the inventory, gets hardware priority from Nvidia, and is first to market with Rubin systems. That is a strong position in a market with no spare capacity.
GTC status
First Rubin provider
Rubin launch
H2 2026
Business model
GPU-as-a-Service
Nvidia relationship
Equity + hardware priority

The Bigger Picture

GTC 2026 showed that Nvidia is building more than chips. Huang wants Nvidia to become the operating system for the age of AI agents. That means software tools, networking products, storage systems, and robotics platforms — on top of the GPUs.

If even part of that vision works, the $1 trillion demand forecast becomes a floor — not a ceiling. Autonomous vehicles, industrial robots, and AI data centres all need Nvidia. That is a decade of growth, not just a two-year cycle.

For investors, the question is not whether to have AI exposure. It is how to spread that exposure across the supply chain. TSMC, Marvell, Microsoft, SK Hynix, and CoreWeave each cover a different layer. Together, they let you participate in Huang’s $1 trillion buildout without relying on any single company to execute perfectly.

⚠️ Disclaimer — Analysis, Not Advice

This article is for information only. It is not personal financial advice. All investments carry risk, including the risk of losing money. Past performance does not predict future results. Please do your own research and speak to a licensed financial adviser before making any investment decisions. investnotbet.com does not hold positions in any stocks mentioned at time of publication.

Nvidia NVDA GTC 2026 Jensen Huang Vera Rubin TSMC Marvell CoreWeave SK Hynix Microsoft Agentic AI AI Infrastructure Semiconductors

Leave a Reply

Powered by WordPress.com.

Up ↑

Discover more from

Subscribe now to keep reading and get access to the full archive.

Continue reading