{"id":512827,"date":"2026-03-09T10:53:13","date_gmt":"2026-03-09T10:53:13","guid":{"rendered":"https:\/\/www.newsbeep.com\/us\/512827\/"},"modified":"2026-03-09T10:53:13","modified_gmt":"2026-03-09T10:53:13","slug":"this-ai-startup-demands-amd-to-build-a-96-gb-rdna-5-gpu-for-a-wild-venture-and-is-already-seeking-investors","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/us\/512827\/","title":{"rendered":"This AI Startup Demands AMD to Build a 96 GB RDNA 5 GPU for a Wild Venture, and Is Already Seeking Investors"},"content":{"rendered":"

The demand for compute power has gone wild out there, and in the midst of it, the AI startup TinyCorp has a rather interesting demand for AMD, involving a ‘chonky’ RDNA 5 GPU.<\/p>\n

TinyCorp Hopes To Get an RDNA 5 GPU With 96 GB VRAM For $2,500 Each; Probably Ignoring the Memory Shortages<\/p>\n

Every AI startup whose world revolves around ‘tokenomics’ is hungry for compute power and ready to go to any lengths to get it. One of the primary reasons consumer GPUs saw massive availability constraints was that, for small- and medium-scale AI enterprises, they were a viable option for getting decent training\/inference capabilities onboard. And, one of the more prominent AI startups focusing on consumer GPUs was TinyCorp. The startup also got into a fiasco with AMD<\/a>, but for now, it demands that Team Red go “all-in” on RDNA 5 GPUs.<\/p>\n

if tiny corp was raising $20M (@ $200M), who’d be interested? business model is basically this.<\/p>\n

buy this $11.5M building (with 5MW of power): link in our discord<\/p>\n

wait for AMD to launch the RDNA5 96GB cards (mid 2027). preorder 3000 cards (hopefully we can negotiate for $2500\u2026<\/p>\n

\u2014 the tiny corp (@__tinygrad__) March 7, 2026<\/a><\/p>\n

TinyCorp has suggested a pitch to investors that revolves around selling compute power on platforms like OpenRouter using AMD’s consumer GPUs. In a fantasy world, the startup wants to raise $11.5 million to set up a 5MW facility in Oregon, claiming that the region offers better economics for a token sale. Using next-gen RDNA 5 GPUs<\/a>, TinyCorp aims to build a portfolio of 3,000 units and, with that, hopes to generate up to $5.4 million in revenue as a token-selling business at OpenRouter. While the model does look fancy due to its business prospects, it all depends on a ‘flawed’ idea: that AMD would actually release 96 GB RDNA 5 SKUs.<\/p>\n

We do know that RDNA 5 would debut by mid-2027<\/a>, according to the latest information, and while VRAM capacity has been a debate with AMD GPUs ever since the release of the RX 9000 series, scaling up to a 96 GB configuration for a consumer-grade option would be impossible. And when you factor in the memory shortages and the massive supply constraints on general-purpose DRAM products, the idea becomes untenable. Down the line, the only way we see AMD actually scaling up to 96 GB capacity with RDNA 5 could be through its workstation ‘Radeon PRO’ offerings, but yet again, this is just a theory.<\/p>\n

\t\t\t\t\t
\n\t\t\t\t\t\t
\n\t\t\t\t\t<\/a><\/p>\n

\t\t\t\t
\n\t\t\t\t\t
\n\t\t\t\t<\/a><\/p>\n

.@AMD<\/a> @LisaSu<\/a> would be dumb not to make one, I trust they are. RDNA5 should be GDDR7 with a 512-bit bus. get the 3 GB modules and double side it like the Blackwell. we’ll build the board if they don’t \ud83d\ude42<\/p>\n

\u2014 the tiny corp (@__tinygrad__) March 7, 2026<\/a><\/p>\n

Interestingly, TinyCorp claims it will build its own board with RDNA 5 silicon onboard, provided AMD doesn’t release a similar model. The only GPU with a 96 GB capacity is NVIDIA’s RTX PRO 6000 Ada Blackwell. Just for information, a single unit currently retails for around $8,000 to $10,000. TinyCorp’s idea focuses on the fact that the RDNA 5 GPU with staggering VRAM figures would cost $2,500 each, which could never be the case given the world we live in. Anyways, full marks to the idea, as the token business is appealing with consumer GPUs.<\/p>\n

\n\t\t\t\tFollow Wccftech on Google<\/a> to get more of our news coverage in your feeds.\n\t\t\t<\/p>\n

\t\t