AI News August 18, 2026: Nvidia Backstops OpenAI's $105B Ohio Megacampus, Groq Raises at Half Its Valuation, Amazon Cuts Up Rare Books for AI Training, and a ChatGPT-Written Report Just Cost 3M $61M

📰 Top 5 AI Stories — August 18, 2026

The money story today is enormous, and it arrived in an SEC filing. Nvidia has put its balance sheet behind OpenAI’s Ohio megacampus — up to $105 billion in guarantees, the largest data-center commitment ever recorded. Meanwhile, Groq re-raised at half its former valuation, Amazon’s AI training pipeline ran into a rare-books scandal, and a jury made an example of an expert witness who let ChatGPT do his job. Here’s what matters today.

1. 🏗️ Nvidia Backstops OpenAI’s $105B, 10-Gigawatt Ohio Megacampus

The biggest infrastructure deal in AI history just got papered. An August 17 SEC filing shows Nvidia agreeing to guarantee up to $105 billion in lease and power obligations at a 10-gigawatt data center campus in Pike County, Ohio that OpenAI has leased for 20 years, with the first 800MW slated to come online by 2028.

The structure: SoftBank subsidiary SB Energy will build, own, and operate the campus on a decommissioned uranium-enrichment site about 50 miles south of Columbus. Nvidia is also investing $1.5 billion in SB Energy, will exclusively supply the campus’s AI compute, and holds an option on the remaining 3.75 GW beyond OpenAI’s initial 4.25-gigawatt commitment. Total project cost including chips could exceed $500 billion.

The guarantee is Nvidia’s biggest financing commitment ever — larger than its prior CoreWeave backstops — though notably smaller than the $250 billion originally discussed in WSJ’s reporting. It’s the sharpest example yet of “circular financing” concerns: the chip vendor guaranteeing its customer’s leases so the customer can buy the vendor’s chips. CNBC notes the deal lands just a week after Nvidia lined up $500 billion in customer financing for hardware purchases. When one company both supplies the compute and underwrites the debt, the AI buildout’s risk is concentrating in exactly one place. (Sources: Bloomberg, CNBC, Reuters)

2. 📉 Groq Raises $350M at $3.5B — Half Its Prior Valuation

Inference-chip startup Groq raised $350 million led by Disruptive at a $3.5 billion valuation — roughly half the $6.9 billion mark it hit last September, per Bloomberg.

The context is bruising: months ago, Nvidia struck a licensing deal that hired away founder-CEO Jonathan Ross and much of the team. With Nvidia now participating in the round, Groq is remaking itself as an inference-focused data-center operator, aiming to expand beyond 200MW of capacity next year.

The read: the window for standalone AI-chip challengers has narrowed dramatically. Groq survived — with capital and a pivot — but a 50% down round for one of the sector’s best-known hardware startups says plainly that investors now see inference acceleration consolidating around the incumbent. (Source: Bloomberg)

3. 📚 Amazon Is Cutting Up Rare Books to Scan Them for AI Training

An investigation by 404 Media placed an AirTag in a box of roughly 1,000 rare books and traced it from California through Milwaukee, a Kenosha warehouse, and a Colorado truck stop to Amazon’s LAS8 facility in Las Vegas — home of the VGT3 operation dedicated to book scanning.

Employees described the workflow plainly: workers cut off the spines and scan barcodes and pages to digitize the contents. Amazon confirmed it “purchases books through commercial channels to help develop and improve the products and services our customers use” but declined to discuss methods or scale.

Why it matters: this is the first documentary evidence of a hyperscaler physically destroying physical books as an input pipeline for AI training data. Expect this story to surface in the ongoing publisher lawsuits over training-data provenance — and in state legislatures weighing AI training-data disclosure bills. (Source: 404 Media)

4. ⚖️ 3M Expert Had ChatGPT Write 90% of His Report — and Lost a $61M Verdict

Court records unsealed in the Watson Grinding lawsuit show expert witness Josh Autenrieth of Knighthawk Engineering prompted ChatGPT to “show how 3M is 0% at fault” and produced an expert report that trial testimony pegged at 85–90% ChatGPT-generated. Discovery yielded 350 pages of prompts.

The jury’s answer: a $61 million verdict, assigning 30% of the fault to 3M, in a case over a deadly industrial explosion.

This is now the marquee example of AI-generated expert testimony being discovered and used to impeach a witness. The lesson for every professional who touches AI in high-stakes work: prompts are discoverable, AI authorship is quantifiable, and juries punish it. Your ChatGPT thread is not privileged, and it is not a draft — it’s evidence. (Source: 404 Media)

5. 📊 Quick Hits: $3T Off-Balance-Sheet, Copilot Autofix Exploited, Robot Dogs Fight Fires

• Big Tech’s real AI bill is $3 trillion. A WSJ analysis found nine top tech firms — Alphabet, Amazon, Meta, Microsoft, Oracle, Nvidia, Broadcom, SpaceX, and AMD — carry roughly $3 trillion in AI-related off-balance-sheet commitments versus about $600 billion in reported 12-month capex. Alphabet alone lists $811B in purchase and contractual commitments. Michael Burry, short Oracle and Nebius, argues hyperscalers use 5–6 year chip depreciation schedules when the economics say 2–3. (Source: WSJ)
• A GitHub Copilot Autofix patch created a real vulnerability. Wiz’s Red Team found Copilot Autofix replaced a safe input pattern with raw string interpolation of a GitHub issue title in Snowflake’s connector repo — and a flawed conditional gate let an unauthenticated attacker exploit it within five days, exfiltrating a Jira token. AI-suggested patches now need the same review as human ones. (Source: Wiz)
• Chinese robot dogs are being sent into fires ahead of humans. DEEP Robotics is fielding its X30 quadruped with a 60m water cannon, 3D LiDAR, and multi-gas sensors into high-rise fires, tunnel emergencies, and hazmat scenes — forming mesh networks to relay data from low-signal environments. (Source: Interesting Engineering)

🔮 The Takeaway

Today’s stories share one spine: the bill for the AI boom is coming due in places nobody budgeted for. Nvidia’s $105B guarantee puts the chipmaker’s own credit on the line to keep the buildout moving — the same buildout WSJ now totals at $3 trillion in off-balance-sheet promises. Groq’s down round shows what happens to challengers when that consolidation accelerates. Amazon’s book-cutting operation and the 3M fiasco are the same story at different scales: institutions feeding the machine faster than they’re accounting for the consequences — whether the cost lands on copyright law, a jury verdict, or a warehouse of rare books that no longer exist.

Cover image generated with Cloudflare Workers AI (SDXL). Sources linked above; stories summarized from Bloomberg, CNBC, Reuters, 404 Media, WSJ, Wiz, and Interesting Engineering.