Google opened day two of I/O 2026 with an announcement that reshaped the AI industry conversation: Gemini Omni Flash, the first of a new family of multimodal world models that can generate video, understand physics, perform natural-language edits, and output any format from any combination of inputs. The announcement caps a dramatic week in which the world's three leading AI labs all made major moves.

What Gemini Omni Changes

Gemini Omni Flash is not another large language model. Google is calling it a "world model" — a system trained to simulate physical reality: kinetic energy, gravity, object interactions. It unifies capabilities previously scattered across separate models like Nano, Genie, and Veo, enabling high-quality video generation from any mix of text, images, video, or audio. According to reports from The Verge, the model is available immediately to Google AI Plus, Pro, and Ultra subscribers, and will roll out to YouTube Shorts later this week.

The implication: instead of a model that only predicts the next token, Google is offering a system that understands how the world works and can manipulate it. "This is a paradigm shift, not a version upgrade," one user wrote on social media in response to the announcement.

OpenAI: Trial Win Clears IPO Path

The day before Google I/O, OpenAI secured a major legal victory. A federal jury in Oakland unanimously rejected Elon Musk's $150 billion lawsuit, ruling the claims were barred by the statute of limitations. Musk, a co-founder who left OpenAI in 2018, had argued the company abandoned its original nonprofit mission. The jury deliberated less than two hours before delivering the verdict.

The win removes a heavy legal cloud over OpenAI's IPO plans, according to a Reuters report. Analysts estimate the company's potential valuation at up to $1 trillion, and the path to going public — at least on the legal front — now looks clearer.

Meta: 8,000 Laid Off, 7,000 Shifted to AI

Meta began a wave of layoffs today, cutting roughly 8,000 employees as part of a 10% workforce reduction. The twist: the company is reassigning 7,000 workers to four new AI-focused divisions, concentrating on AI-native tools, applications, and workflows. Internal documents show Meta plans to spend between $115 and $135 billion on AI this year, including compute infrastructure and data centers.

Anthropic: Gates Foundation, SpaceX, and Claude for Small Business

Anthropic, the company behind Claude, had a busy week. It signed a $200 million, four-year partnership with the Bill & Melinda Gates Foundation — one of the largest AI deals in the philanthropic sector. Separately, Anthropic struck a major compute agreement with SpaceX for its Memphis data center and launched "Claude for Small Business," a package with built-in integrations for QuickBooks, HubSpot, and Microsoft 365.

Alibaba Unveils Its Own AI Chip

On the Chinese front, Alibaba revealed the Zhenwu M890 AI accelerator, delivering three times the performance of its predecessor. With 144 GB of GPU memory and 800 GB/s interchip bandwidth, it represents the Chinese tech giant's push to reduce dependence on Nvidia amid ongoing U.S. export restrictions.

The Bottom Line

Today made clear: the AI industry is shifting from a model race to an application race. Google is opening a new front with world models, OpenAI is clearing the decks for an IPO, Meta is repositioning nearly a quarter of its workforce around AI, and Anthropic is building broad enterprise infrastructure. Meanwhile, Alibaba shows that China has no intention of sitting this one out.

The coming weeks will determine which of the big players can translate enormous investments into products average users actually want to pay for.