Today's AI landscape is defined by a striking contradiction: capital and ambition are accelerating even as evidence of real-world returns remains elusive. OpenAI is reportedly assembling a $30 billion raise at a $1.25 trillion valuation ahead of an IPO, while an Apollo economist argues most companies are seeing "zero financial payoff" from AI. Meanwhile, OpenAI has made a landmark decision to stop training on the live internet, Grok and OpenAI's models display divergent political leanings under forced choice, and the agent economy takes shape through new authentication tools and text-message-native assistants. Below are the stories that matter most.
OpenAI is reportedly working on a new funding round of roughly $30 billion at a staggering $1.25 trillion valuation, a move that would make it one of the most valuable private companies in history and set the stage for a public listing. The scale of the raise signals that the company intends to keep spending aggressively on compute, talent, and infrastructure despite persistent questions about profitability across the sector. If completed, the round would further concentrate market power in a handful of frontier labs and raise the stakes for competitors racing toward their own public-market debuts.
Source: TechCrunch AI
In one of the most consequential policy shifts of the year, OpenAI says it will no longer train its AI models on the live internet, citing "significant risks." The decision marks a departure from the web-scraping paradigm that underpinned the first generation of large language models and could reshape how frontier labs source data going forward. It also raises immediate questions about data provenance, licensing deals, and whether competitors will follow suit or treat live-web training as a competitive advantage.
Source: TechCrunch AI
A top economist at Apollo is throwing cold water on the AI productivity narrative, arguing that most companies are seeing "zero financial payoff" from their AI investments. The assessment cuts against the torrent of enterprise AI spending and vendor promises, and echoes growing skepticism about whether current-generation tools translate into measurable bottom-line gains. For CFOs and boards, the remark lands as a timely caution: adoption metrics and pilot programs are not the same as returns.
Source: TechCrunch AI
Grok has released the first seven minutes of its AI-generated "Odyssey," a high-profile showcase of what generative video can produce at feature-film scale. The release positions the model as a direct challenger in the increasingly crowded AI filmmaking space and will invite scrutiny over visual consistency, narrative coherence, and the copyright questions that accompany AI-generated cinema. It is also a marketing play: a seven-minute proof point is far more persuasive to studios and creators than a benchmark score.
Source: TechCrunch AI
Block has unveiled a new voice model that reviewers describe as sounding "eerily human," pushing the quality bar for synthetic speech another notch higher. The model arrives as voice interfaces become a primary battleground for AI assistants, customer service, and accessibility tools. The bigger story is the shrinking gap between synthetic and real audio — and the fraud, consent, and disclosure challenges that come with it.
Source: TechCrunch AI
A comparison test found that when forced to choose, Grok's AI model leans Republican while OpenAI's leans Democratic — a vivid illustration of how training choices, data, and alignment decisions can produce divergent political behavior. The findings will intensify scrutiny of model bias as AI systems increasingly mediate news, search, and civic information. For enterprises deploying these models, the results underscore the need for evaluation and guardrails rather than assuming neutrality.
Source: TechCrunch AI
1Password has launched a capability that allows AI agents to authenticate on a user's behalf without ever exposing the underlying password. It is a foundational piece of infrastructure for the emerging agent economy: if autonomous agents are going to book travel, file expenses, and manage accounts, they need secure, scoped credentials. The move also signals that identity and access management is becoming a critical control layer for agentic AI deployments.
Source: TechCrunch AI
TechCrunch has compiled a running list of AI projects and startups that have shut down, offering a sober counterweight to the hype cycle. The graveyard is a useful reality check for founders and investors: for every headline funding round, there are numerous ventures that ran out of runway, misjudged demand, or were steamrolled by foundation-model providers. It is essential reading for anyone calibrating expectations in the current market.
Source: TechCrunch AI
Uber rival inDrive is planning robotaxi tests in Central Asia by 2027 in partnership with Pony AI and Nvidia, extending the global robotaxi race into new geographies. The tie-up pairs a regional ride-hailing player with autonomous-driving technology and compute from two established suppliers. It reflects a broader pattern: robotaxi expansion is increasingly happening through partnerships in underserved markets rather than only in a handful of U.S. and Chinese cities.
Source: TechCrunch AI
Canva has shipped its first in-house image generation model, bringing generative visuals directly into its design workflow rather than relying solely on third-party APIs. The move is a strategic bet that owning the model improves quality, cost, and integration for its large base of non-designer users. It also intensifies competition with Adobe and a wave of AI-native design tools racing to embed generation into everyday creative work.
Source: TechCrunch AI
The throughline today is a market learning to separate narrative from numbers. OpenAI's reported $1.25 trillion valuation and Grok's cinematic ambitions show that the appetite for frontier AI remains enormous. But the Apollo economist's "zero financial payoff" finding and the AI graveyard remind us that enterprise returns are lagging the hype. The most durable developments may be the unglamorous ones — OpenAI's training-data policy shift, 1Password's agent authentication, and inDrive's robotaxi partnerships — because they set the rules and rails on which the next phase of AI will actually run.