Nvidia (NVDA) and the whole AI story
AI starts and ends with Nvidia (NVDA). What Jensen Huang says on the earnings call, and the company's forecast, ripples across chips, networking, infrastructure, data centers, and software with a 10x scale effect. The entire AI industry rests on his words at this moment.
Data centers and the numbers
Data centers are the good news. NVDA beat estimates in 22 of the last 24 quarters, and 5 quarters in a row. The street expects about $92 billion for data centers; a surprise could push it toward $110 billion. This may be the first $100 billion revenue quarter.
For the outlook and guidance, the forecast needs to land in the $105 to $110 billion range to excite investors.
The backlog is huge: about $700 billion this year, $1.2 to $1.3 trillion in 2027, and potentially $1.6 trillion in 2028. That backlog is the ripple effect. It creates a big opening for anyone building high-bandwidth memory, network connectors, photonics, and server rack systems.
Risks and concerns
Possible problems: politics blocking data center construction, restrictions on selling chips to China, and shortages of high-bandwidth memory chips. NVDA is raising prices because of the costs it faces, and the market has read those price hikes as good for tech overall. If there is a memory chip shortage, supply chain bottlenecks, or regulatory hurdles, it matters because everything connects back to Nvidia (NVDA) - the company sits in the middle like the sun, with everything else orbiting it.
Blackwell, Rubin, and the road map
Blackwell is still selling well, with many orders coming in. Investors want to gauge ongoing demand for both Blackwell and Rubin. The genius move was the road map. Chip makers usually publish plans no more than two generations ahead; Huang laid out the next five, even the next ten generations. No matter which chip you start with, you can always swap it for a newer one. Once you are on Nvidia (NVDA), you stay with Nvidia.
Software: CUDA and Nemotron
Everything comes to life in software. CUDA and Nemotron give enterprises the enterprise-grade software they need, gaining traction even against OpenAI or Anthropic. The partnerships and deals on these products are impressive; most are not public, but partners and vendors confirm them, and enough public ones show real traction.
When will this software growth show up? The partnerships are happening today. Steady growth builds over the next 8 to 12 quarters, then a hockey stick kicks up - driven by the right chips, hardware, networking, and use cases. Think of it in stages: year one is partnerships (software vendors, hardware vendors, industrial companies working on Nemotron); revenue starts picking up 24 months out, and the hockey stick hits around 36 months. Adoption is happening now, but the revenue is not yet obvious.
Bigger markets ahead
Midterm growth comes from physical AI and sovereign AI, each a trillion-dollar market. Growth is also expanding beyond GPUs into CPUs, which is climbing fast, plus all AI-driven areas.
Chips as a new asset class and the financing move
Chips have become a new asset class, like crypto, bonds, or gold. At both the start and the end, it always comes down to a token, and a token needs a chip. Nvidia (NVDA) has turned that into an asset class.
Funding has shifted from cyclical, inside-the-industry money to real institutions. Six of them have put together about $500 billion. Other companies are starting to copy the idea, though not at the same scale. Huang probably will not comment on this or give details on the call.
Look at Nvidia's (NVDA) partnerships and investment strategy over the past 12 months and you see bets across every part of the six-layer cake: startups and the wider ecosystem. The company acts as a good neighbor, making sure everything and everyone is in the right place and set up to succeed. In tech history, one group rarely does well while every connected group also does well; that happens here because of Huang's stewardship, and he deserves credit for it.
Competition and supply
There are alternatives: CPUs and TPUs. Every hyperscaler is building its own chip because they feel they must. Google (GOOGL) TPUs are in-house, and some are looking at recent OpenAI announcements. Ask a chip engineer and Nvidia's (NVDA) lead is three to four generations ahead of the nearest competitor. The question now is cost per performance. NVDA may lose a little market share to TPUs and others, but overall demand is 10x bigger than most people think. There are still not enough chips; supply is massively constrained by the ability to make enough chips for the use cases in play.
Valuation
You cannot compare today's AI to the 1999-2000 tech bubble, and Nvidia's (NVDA) valuation has actually come down.
Stock and options
The stock trades around $210. The options market prices a 5.5% swing. Most trades stay bullish, with little positioning below the $200 mark. The $220 strike is very popular, with heavy call volume.
