Nvidia demand spreads

Nvidia’s latest message is that artificial intelligence demand is no longer just a chip story. It is spreading into servers, data centers and local computing services built around its Blackwell systems. (cnbc.com) At Nvidia’s GTC conference on March 16, 2026, Chief Executive Jensen Huang said he sees $1 trillion in orders for Blackwell and Vera Rubin systems through 2027. That was up from the roughly $500 billion opportunity he cited a year earlier through 2026. (cnbc.com) A graphics processing unit, or graphics chip, is the part that does the heavy math for training and running artificial intelligence models. Blackwell is Nvidia’s current data-center generation, and Vera Rubin is the next one in line. (cnbc.com) That demand is showing up in the companies that package Nvidia chips into usable machines. Super Micro said on February 3 that fiscal second-quarter 2026 net sales reached $12.7 billion, and it forecast at least $40.0 billion in sales for the full fiscal year. (ir.supermicro.com) Super Micro has been shifting toward complete artificial intelligence systems rather than selling parts one by one. Its investor materials for the February 3 results highlighted “total IT/AI” systems, and outside market summaries of the same quarter said artificial intelligence infrastructure made up more than 90% of revenue. (ir.supermicro.com, investing.com) The spread is also geographic. On April 14, HostAfrica said it launched a local graphics-chip service from Digital Parks Africa’s Samrand data center using Nvidia RTX 6000 Blackwell Server Edition processors with 96 gigabytes of GDDR7 memory. (it-online.co.za) HostAfrica said the service is aimed at artificial intelligence, machine learning, rendering and other data-heavy jobs that companies want to run inside South Africa. Digital Parks Africa said local hosting can reduce delay and keep sensitive data within national borders. (it-online.co.za) That local angle matters because many companies do not buy chips outright. They rent computing time, which turns Nvidia demand into recurring business for server makers, cloud providers and regional data-center operators. (it-online.co.za, ir.supermicro.com) Nvidia is also using more artificial intelligence inside its own engineering process. Tom’s Hardware reported on April 13 that Nvidia said a chip-design task that once took eight engineers 10 months can now be done overnight with one graphics processor, though the company said fully autonomous chip design is still far off. (tomshardware.com) The result is a wider artificial intelligence supply chain than the market had a year ago. Nvidia still sits at the center, but the money is now flowing through rack builders, liquid-cooling specialists, hosting firms and local cloud operators that can turn scarce graphics chips into usable capacity. (cnbc.com, ir.supermicro.com, it-online.co.za)