OpenAI Slashes Prices Ahead of IPO, Microsoft Drops Third-Party AI, Chinese Models Close the Gap — AI News Briefing

Top 7 Stories

1. OpenAI and Anthropic Wage Token Price War as IPO Race Heats Up

The two AI giants are handing out millions of dollars in free computing credits to attract startup customers, with the WSJ reporting that OpenAI is considering drastic price cuts in anticipation of an all-out war for users with Anthropic. Both companies are speeding toward IPOs in what analysts are calling a $3.8 trillion “mega IPO wave,” but their road is getting bumpier as Wall Street scrambles to understand the token-based revenue model.

Palantir CEO Alex Karp didn’t mince words, calling the industry “effing insane” and slamming both companies’ token pricing as fundamentally broken. “Something has gone completely wrong,” Karp said, arguing that the race-to-the-bottom pricing will make it difficult for either company to demonstrate sustainable economics to public market investors. The Financial Times echoed the concern, questioning whether either company can float successfully given growing scrutiny of their business models.

2. Microsoft Replaces OpenAI and Anthropic with In-House Models in Excel and Outlook

In a significant shift, Microsoft has begun replacing OpenAI and Anthropic models with its own proprietary AI in core Office products including Excel and Outlook. The move, reported by TechCrunch and the Hindustan Times, marks a dramatic turn for a company that was once OpenAI’s biggest backer and most important distribution partner.

The transition reflects a broader industry cost-cutting trend as enterprises question the economics of relying on expensive third-party API calls. Microsoft’s decision to vertically integrate its AI stack mirrors what Google and Meta have done from the start, and raises serious questions about the long-term durability of OpenAI’s enterprise revenue — particularly as the company races toward an IPO.

3. Chinese AI Models Z.ai and GLM-5.2 Close the Gap with Western Leaders

A new generation of Chinese AI models is rapidly approaching parity with OpenAI and Anthropic on their home turf. Z.ai, a Chinese startup, has emerged as a direct challenger with an inexpensive model that Reuters reports is “catching up with Anthropic and OpenAI.” Meanwhile, GLM-5.2 from Tsinghua-backed Zhipu AI is demonstrating competitive performance at a fraction of the cost.

The narrowing gap has turned AI distillation into a geopolitical flashpoint. The New York Times reports that distillation — the technique of training smaller models on outputs from larger ones — has become a central issue in the US-China AI race, with Washington increasingly concerned that export controls on chips alone won’t be enough to maintain American dominance.

4. Anthropic’s Claude Cowork Expands to Mobile and Web

Anthropic launched Claude Cowork on mobile and web platforms this week, extending its collaborative AI workspace beyond the desktop. The expansion, first reported by TechCrunch, brings Anthropic’s agentic coding and collaboration features to a much broader audience, allowing users to work alongside Claude on projects from any device.

The move positions Anthropic to compete more directly with GitHub Copilot’s multi-platform strategy and signals that the company sees agentic collaboration — not just chat — as the future of human-AI interaction. The timing is notable as the company prepares for its own IPO, with investors watching closely for signs of product differentiation and user growth.

5. Discord AI Moderation Wrongfully Bans Users Over Harmless Images

Discord admitted this week that an AI moderation bug wrongfully banned users for sharing harmless images, reigniting the debate over automated content moderation at scale. The platform acknowledged the error after users reported being suspended for completely innocuous content, highlighting the persistent challenge of false positives in AI moderation systems.

The incident underscores a growing tension as platforms deploy increasingly aggressive AI moderation to comply with regulatory pressure while users push back against what they see as overreach. With the EU’s Digital Services Act and similar regulations demanding proactive content removal, the margin for error is shrinking — and users are paying the price when algorithms get it wrong.

6. NVIDIA Taps Cloud Providers to Expand Startup Compute Access

NVIDIA has launched a new initiative partnering with AI cloud providers to expand compute access for startups, CNBC reports. The program offers startups discounted access to GPU clusters through NVIDIA’s cloud partners, complementing the free token giveaways from OpenAI and Anthropic and effectively making NVIDIA the arms dealer in the AI platform war.

The move comes as NVIDIA rides blistering chip sales to another record quarter — $81.6 billion — with data center revenue surging 92% year-over-year. Jensen Huang is also eyeing the CPU market, arguing that the rise of agentic AI will reshape chip demand beyond GPUs alone.

7. AI PACs Pour $20 Million Into Elections as Voters Demand Tougher Regulation

AI industry PACs have poured over $20 million into the New York Democratic primary in what CNBC and The Hill describe as a proxy war between the AI industry and safety advocacy groups. The flood of money underscores how AI regulation has become one of the most contentious political battlegrounds of the 2026 election cycle — and it’s not just industry spending: safety-focused groups are mobilizing their own war chests.

A new NBC News poll found that voters of both parties overwhelmingly want tighter AI regulation, creating an awkward dynamic for candidates who accept industry money. The bipartisan consensus on regulation is striking in an otherwise polarized political environment, suggesting that whoever wins the policy battle will shape the industry for a generation.

Trend Watch

What to Watch

OpenAI and Anthropic IPO filings. With both companies expected to file S-1s in the coming weeks, the disclosures will reveal — for the first time — real revenue, burn rates, and customer concentration risk. Wall Street’s ability to value token-based businesses will be tested like never before.

Microsoft Build and Google I/O fallouts. As both tech giants double down on in-house models, watch for announcements that further marginalize third-party API providers. The “model as a commodity” thesis is gaining momentum.

EU AI Act enforcement begins. Implementation of the EU’s landmark AI law is accelerating, with the first major enforcement actions expected this summer. Companies relying heavily on AI moderation and automated decision-making face the highest exposure.

China’s next model drop. After Z.ai and GLM-5.2, the AI world is waiting for the next Chinese frontier model. If the trend of closing the gap continues, expect Washington to respond with new restrictions beyond chip export controls.