AI News September 2, 2026: ChatGPT Ads Hit $1B Run Rate, Nvidia Parks $3.5B in MediaTek, Google Poaches a Star From OpenAI

September opens with the AI economy’s money stories front and center: OpenAI has proven it can monetize attention at scale, Nvidia is hedging the custom-chip threat by owning the plumbing, Google keeps buying star researchers, and both the IPO window and the US–China diplomatic channel open this month.

1. ChatGPT Ads Hits $1 Billion Run Rate in Under 200 Days

OpenAI announced Monday that its advertising business has reached a $1 billion annualized revenue run rate — less than 200 days after ChatGPT Ads launched, CNBC and Reuters report. The platform now serves tens of thousands of advertisers across more than 40 countries, and self-service buying via Ads Manager is rolling out across India, Europe, the Middle East and North Africa.

The company emphasized guardrails: ads are clearly labeled, never influence chatbot answers, and advertisers cannot access private conversations. Ads run for free-tier and Go subscribers — the vast majority of ChatGPT’s reported 1 billion weekly users.

Why it matters: A billion-dollar ads business alongside a $40 billion-plus overall run rate gives OpenAI a genuine “diversified business model” story exactly as it prepares to go public. It also drags the search-ad economy deeper into conversational AI — a direct threat to Google’s core franchise, and a sign that the free-tier AI wars will be ad-funded.

2. Nvidia Puts $3.5 Billion Into MediaTek to Own the Custom-Chip Era

Nvidia is investing $3.5 billion in Taiwanese chipmaker MediaTek. As part of the deal, MediaTek will adopt Nvidia’s technology to design custom chips for AI companies and hyperscalers that plug directly into Nvidia-based data centers, TechCrunch reports.

The logic: Amazon, Google, Microsoft, OpenAI and Anthropic are all building their own XPUs to escape Nvidia GPU dependency. Rather than fight the trend, Jensen Huang is making sure those custom chips still connect into Nvidia’s AI-factory architecture. “We are not intimidated by this — we are in fact welcoming it,” Huang told Bloomberg. The companies will also keep collaborating on DGX Spark and RTX Spark consumer AI PCs.

Why it matters: This is Nvidia’s insurance policy. Even if Big Tech’s custom silicon takes share, Nvidia collects tolls on the interconnect, networking and software layers. Combined with the reported $12.9 billion Hugging Face deal and this week’s 17% flagship chip price increase, Nvidia is methodically converting GPU dominance into full-stack leverage.

3. Barret Zoph Defects From OpenAI to Google DeepMind

Barret Zoph — Thinking Machines Lab co-founder, former OpenAI post-training lead, and famous for the falling-out with CEO Mira Murati — has changed employers yet again, joining Google DeepMind as Vice President of Research, the WSJ and Reuters report. Google confirmed the hire, saying it looks forward to Zoph “bringing his RL and post-training expertise to Gemini.”

Zoph spent six years at Google Brain before following OpenAI’s post-training team culture into Thinking Machines in 2025, returning to OpenAI in January, and leaving again in June. He lands at a DeepMind in transition, following an August reorganization that moved some teams out of the lab into corporate Google.

Why it matters: Post-training and reinforcement learning are where frontier-model quality is won in 2026. Google — which has watched OpenAI and Anthropic poach its researchers for years — is now winning a marquee name back, and betting him on Gemini.

4. Anthropic’s IPO Prospectus Lands Right After Labor Day

The Information reports Anthropic plans to publicly unveil its IPO prospectus just after Labor Day, targeting a listing in late September or early October — potentially surpassing SpaceX’s record $86.2 billion raise. Unusually, Anthropic is weighing letting existing shareholders sell shares in the offering, with lockups possibly longer than the standard 180 days. Prediction markets now price a real chance of a valuation above $2 trillion.

Why it matters: The public filing will force Anthropic’s financials into the open — revenue run rate, compute commitments (including that $45 billion Nscale deal), and losses. It sets the template for OpenAI’s own debut, and for how public markets will value AI labs for years.

5. US and China Hold First Official AI Talks This Month

Reuters reports the United States and China will hold their first formal AI dialogue of the Trump administration in September, led by Treasury Secretary Scott Bessent, ahead of Xi Jinping’s September 24 visit to the US. The agenda covers frontier-model risks: military capabilities, cyberattacks on critical infrastructure, and labor disruption.

Why it matters: AI now has its own diplomatic lane. With US chip export controls tightening and China championing open-weight models globally, whatever comes out of these talks will shape how models, chips and research actually flow between the two AI superpowers.

Coverage draws on reporting from CNBC, Reuters, TechCrunch, Bloomberg and The Information. This digest is published daily at AI Tools Hub.