ULTRA Loading

Initializing System

Skip to content
← Back to blog
aitechroundupnvidiabitcoin-mining

AI & Tech News Roundup — May 15, 2026

Cerebras goes public in the biggest AI IPO of 2026, NVIDIA commits $40B to equity deals including a $2.1B stake in Bitcoin miner IREN, and OpenAI shuts down Sora after six months.

May 16, 20265 min readBy Ultra Labs
AI & Tech News Roundup — May 15, 2026

AI & Tech News Roundup — May 15, 2026

The AI infrastructure buildout is producing increasingly strange contradictions: the biggest tech IPO of 2026 is a chip company, NVIDIA is investing billions directly into data center operators (including a Bitcoin miner), and OpenAI just killed its flagship video product because it burned $15 million a day in compute costs. This week had all three.


1. Cerebras Pops 68% in the Biggest AI IPO of 2026

Wafer-scale chipmaker Cerebras priced its Nasdaq IPO at $185 a share on Wednesday, raised $5.55 billion — the largest US tech IPO since Uber in 2019 — and closed its first trading day up 68% at $311, pushing its market cap to approximately $95 billion.

The Cerebras WSE-3 is a single wafer-sized chip roughly the size of a dinner plate, compared to the thumbnail-sized GPU dies NVIDIA stacks into its H100 and Blackwell systems. The architecture eliminates inter-chip communication bottlenecks that slow down large model inference — which matters enormously for the real-time AI applications that hyperscalers are now trying to run at scale.

Cerebras has an existing $10 billion contract with OpenAI and $510 million in annual revenue, but the $95 billion valuation — roughly 186x revenue — reflects bets on a future where wafer-scale inference becomes the preferred architecture for frontier model deployment. The market clearly believes that future is credible.

For the compute infrastructure thesis that underpins a lot of our Bitcoin miner pivot-to-AI coverage, the Cerebras debut is a useful data point: the market is willing to pay enormous premiums for hardware that solves the data center density and bandwidth bottleneck problems that GPU stacks struggle with.


2. NVIDIA's $40B Equity Spree: Corning, IREN, and the Infrastructure Moat Strategy

NVIDIA has committed more than $40 billion to equity investments in AI companies in the first half of 2026 — a figure that reflects a deliberate strategy to own pieces of the physical infrastructure its chips require to operate.

Two deals stand out. First, NVIDIA agreed to invest up to $3.2 billion in Corning, the glass manufacturer, as part of a push to secure the specialty glass substrates used in advanced chip packaging and data center cabling. Second — and directly relevant to the mining-to-AI infrastructure story — NVIDIA signed a deal for the right to invest up to $2.1 billion in IREN, the data center operator formerly known as Iris Energy, one of the largest publicly listed Bitcoin miners that pivoted aggressively toward high-performance computing infrastructure.

The IREN investment is significant for several reasons. It validates the thesis that Bitcoin mining facilities — built for high-density power delivery, precision cooling, and 24/7 operations — are genuinely competitive starting points for AI data centers. It also gives IREN access to NVIDIA's supply chain and engineering relationships at a time when GPU allocation is still constrained. And it sets a precedent: NVIDIA is now directly invested in what was, until recently, a Bitcoin mining company.

Separately, NVIDIA and OpenAI announced a 10-gigawatt deployment partnership for NVIDIA's Vera Rubin platform, with NVIDIA intending to invest up to $100 billion in OpenAI progressively as capacity deploys.


3. OpenAI Kills Sora: $15 Million a Day Burns Too Fast

OpenAI announced it is shutting down Sora, its AI video generation app, just six months after its public launch. The product reached over a million downloads in its first week. It then fell to under 500,000 active users while burning an estimated $15 million per day in compute costs.

The math on generative video is brutal compared to text. Generating a 10-second video clip at high quality requires orders of magnitude more compute than a conversational exchange. At current GPU pricing, consumer-priced video generation products either hemorrhage money or compress quality to the point of irrelevance. Sora fell into the first category.

The shutdown illustrates a tension that the entire AI industry is navigating: the products that generate the most user excitement often require the most compute, while the products that generate sustainable revenue (enterprise text, code generation, API inference) are less visually striking. OpenAI is refocusing. The $10 billion NVIDIA partnership and the DeployCo consulting venture suggest where that focus is actually going.


4. Public Opposition to AI Data Centers: 71% Say Not in My Backyard

A Gallup survey found that 71% of US adults oppose having an AI data center in their local area, with 48% strongly opposed. The top cited concerns are water usage, noise, energy consumption, and grid strain.

This is worth tracking for the mining and compute infrastructure world for a simple reason: siting new data center capacity is becoming politically as difficult as it is technically. The municipalities that already have industrial power infrastructure — substations, high-voltage transmission access, zoning for heavy electrical load — are increasingly valuable real estate for data center developers who cannot afford a three-year permitting fight in a suburb that doesn't want them.

Bitcoin mining facilities, which already built political and regulatory relationships around high-power industrial sites, have an underappreciated siting advantage for the AI data center buildout. The NVIDIA-IREN deal is partly a bet on exactly that.


Sources: CNBC — Cerebras IPO debut · CNBC — NVIDIA $40B equity investments · NVIDIA Newsroom — OpenAI 10GW partnership · MarketingProfs — AI Update May 15 · DataCenter Dynamics — OpenAI-NVIDIA deal details