NVDA’s latest quarter revenue was $96.2 billion; $89 billion came from data centers, over 92% of the total, while PCs, game consoles, robots, and automotive were under 8%.
A 1-gigawatt AI factory costs about $50 billion in Jensen Huang’s GTC Berlin keynote, while NVDA’s blog puts it closer to $60 billion; that implies about $50 to $60 per watt of capacity to build.
Only part of that $50-$60 per watt goes to NVDA; the rest pays for land, buildings, power, and cooling.
If a 1-gigawatt data center brings in $40 billion a year, a 3-month delay costs about $10 billion in lost revenue.
NVDA’s DSX pod is built and tested at the factory, then shipped ready to plug in, cutting deployment from months to weeks; each pod holds seven racks of Vera Rubins, 504 GPUs, and about 25 exaflops of AI compute.
DSX pods are all liquid cooled and connected by over 10 miles of cable, with first shipments scheduled early next year.
VRT is named as a power and cooling partner on two NVDA partner lists from GTC Berlin; last quarter sales rose 24% year over year to $3.3 billion and full-year revenue guidance was raised to about $14 billion.
COR provides networking for DSX pods via NVDA Spectrum X Ethernet and new switches using co-packaged optics; a co-packaged optics switch was shown on stage and described as already in full-volume production.
NVDA invested $2 billion in COR in March and signed a multi-billion-dollar purchase agreement; COR revenue rose 34% year over year to just over $2 billion last quarter and trades around 22x forward earnings, below Alphabet, Amazon, and Microsoft.
VRT reports earnings on October 28 and COR reports on November 4.
Liquid is described as the “everything exchange”: perpetual contracts for almost any asset, plus stocks, crypto, commodities, FX, and prediction markets on one platform; it has handled over $6 billion in trading volume since launching last year.
The market structure pitch is 24/7 trading and an AI assistant tied to Quad and Chat GPT for research, trade analysis, and order prep, with no automatic execution and every trade reviewed and confirmed.
There are now over 5 million industrial robots in factories worldwide and more than 600,000 were installed last year; NVDA’s target is to move robot setup off the factory floor and earlier in the life cycle.
Jensen’s robot stack is three computers per robot: one in the data center for training, one for a digital twin, and one inside the robot itself; digital twins are positioned as physically accurate simulations for learning and practice.
Seammens says digital twins can cut factory engineering effort by up to 85% and get new products to market up to 40% faster; CGNX fits that setup with machine vision systems, cameras, and AI software for production-line inspection and guidance.
CGNX launched a Jetson-powered vision controller delivering up to 157 trillion AI operations per second; its One Vision platform is used by hundreds of customers, revenue hit a record $291 million last quarter, EPS jumped 80%, cash and investments were $755 million, and debt was zero.
Jensen said telecom networks have about 100,000 sites across the Western world totaling about 100 gigawatts; AWS added 3.9 gigawatts of power capacity last year, and converting telecom networks into distributed AI capacity is framed as a big opportunity.
AI RAN is described as running cell-site infrastructure as software on NVDA-accelerated computers instead of specialized 5G radio hardware; NVDA, T-Mobile, and Nokia completed the world’s first AI RAN call on a live network at a commercial cell site in Washington this month.
NOK is the telecom-infrastructure pick; NVDA invested $1 billion in NOK stock at $6 per share last year, NOK trades around $8 today, sales to AI and cloud customers more than doubled last quarter, and optical and IP networking is expected to grow 25% or more next year.
NVDA released Alpameo at CES in January as an open-source model family for video-based driving decisions and explanations; it is being used to train smaller self-driving models that run in cars, and Jensen called it the ChatGPT moment for self-driving cars.
NVDA and UBER announced robo-taxi plans on NVDA’s self-driving stack and UBER’s ride-hailing network starting in Los Angeles and San Francisco in the first half of 2027, with expansion to driverless service in 28 cities by 2028.
BYD, Ji, Isuzu, and Nissan are building robo-taxi-ready cars on the Hyperion platform and account for over one out of every nine new vehicles sold on Earth last year; Volvo, Mercedes, and JLR signed on as Hyperion-compatible at GTC Berlin, adding roughly 3% of all new vehicles sold last year.
UBER is called the stock to buy for robo-taxis: it already works with more than 30 self-driving partners, revenue grew 18% last year, operating margins climbed to over 12%, and it trades at 17x forward earnings, less than 10% of TSLA’s 175x multiple.
TSLA is described as having flat revenue last quarter, shrinking operating margins for years, and a 175x forward P/E that is still above the combined forward P/E of AAPL, MSFT, AMZN, GOOGL, META, and NVDA despite slower growth and lower margins.
Humanoid robots are framed as the highest-volume robot form; NVDA and German company Agile Robots built the Agile 1S SN reference robot, about 4.5 feet tall with 31 joints, running on Jetson Thor and Groot.
Jetson Thor is a small in-robot computer on the same Blackwell chip design as NVDA’s data-center GPUs; Groot is NVDA’s open model family for humanoid robots, pitched as a pre-trained robot brain that companies can fine-tune.
The Agile 1S ships early next year preassembled and ready to use, skipping months or years of robot development; Boston Dynamics, Figure, Agility Robotics 1X, and OpenAI are named as working with or evaluating Jetson Thor.
Jetson Thor delivers up to 7.5x the AI compute of Jetson Orin and is 3.5x more energy efficient; the developer kit starts at $3,500.
NVDA is priced at about 17x forward earnings, below the S&P 500’s 19x forward P/E; NOK, UBER, and NVDA are all claimed to trade below that S&P multiple.
The final model allocation is 25% NVDA and 15% split across the other five stocks named in the lineup, with the remaining picks described as NOK, UBER, VRT, COR, and CGNX.
VRT and COR are framed as the AI-factory-buildout names because more than 92% of NVDA revenue comes from data centers; CGNX and NOK are the picks for factory robots and smarter cell networks, while UBER is the network robo-taxis plug into.
Five of the six stocks are said to report earnings in the next few weeks, each is said to have a direct connection to NVDA, and the closing claim is that it does not matter which AI lab, robot maker, or car maker wins because all of them get paid through the stack.
These names are presented as a way to get rich without getting lucky.