The Information AM

Save 25% on an annual subscription to read the most important news about technology and business first. For even more, Save $250 on The Information Pro for unlimited access to our proprietary org charts, databases and surveys.

OpenAI has held early conversations with investors about a new funding round that could lift its valuation to $1.2 trillion or higher, according to people familiar with the conversations. The conversations come as OpenAI has pushed off a planned initial public offering to next year or later.

The potential funding follows interest by some investors in a new round, said one of the people. The company last announced a $122 billion funding round at a valuation of $852 billion, including the new investment, in March. Since then, its estimated value on the secondary market, as tracked by Caplight, has risen past $1 trillion as its Codex coding product gained in popularity and its recently released GPT-6 Astra model outperforms competitors in some areas. The Financial Times first reported on the funding talks.

OpenAI had previously planned an IPO as soon as late this year. But CEO Sam Altman said recently that an offering this year didn’t make sense amid rising concerns about AI safety. Rival Anthropic has been working on a planned IPO expected to be $1.5 trillion or higher.

Meta Platforms is expanding its paid subscription offerings for its AI tools across Instagram, WhatsApp and Facebook as it looks to diversify revenue beyond digital advertising.

The company announced Tuesday in a blog post that it is introducing new tiers of its Meta One subscription plans, allowing users to get higher usage limits of Meta AI than free versions, along with more than 50 features across Meta’s platforms. This includes tools for content creation, audience engagement and business management.

The move gives Meta another way to generate revenue from its massive user base as it looks to reduce its reliance on digital advertising and turn demand for its AI tools into a paid business.

Meta One plans are now available globally, starting at $2.99 a month for individual products such as WhatsApp Plus, $7.99 a month for individual bundles and $14.99 a month for creator and business bundles.

At the higher end, Meta is offering an “Expert” plan for $149 a month for power users of its business and creator bundles. The “Max” plan costs $499 a month and provides the highest levels of feature access, including higher usage access to Meta Business Agent.

Meta previously announced in May that it was testing subscription plans, including Meta One Plus at $7.99 per month and Meta One Premium at $19.99 per month. The service, which Meta says has 15 million subscribers, was initially rolled out in a limited number of countries.

Meta says the new features include AI tools for creating and editing images and videos, Instagram voice effects, enhanced profiles and follow features for creators, and expanded access to Meta Business Agent, which can respond to customers around the clock on WhatsApp. Higher-tier plans also offer more analytics and account management tools. The company said Meta One will soon expand to its smart glasses and its video editing app Edits.

South Korean memory chipmaker SK Hynix is in talks with Intel about a deal to manufacture memory chips on U.S. soil for the first time, Reuters reported, citing three people familiar with the discussions.

SK Hynix would either lease part of Intel’s long-planned Ohio fab, or form a joint venture with Intel and major cloud firms eager to secure memory supplies, according to Reuters.

A tie-up would mark a significant shift for SK Hynix, which has kept its most advanced production at home. The talks come as Washington presses Korean firms to build memory fabs in the U.S. Their new partnership would also expand SK Hynix’s U.S. footprint beyond the $4 billion advanced packaging plant it is already building in Indiana.

The discussions remain preliminary and face steep hurdles, from higher U.S. production costs to regulatory approval in Seoul, according to Reuters. Opposition from Seoul could be the biggest obstacle: any agreement to produce advanced memory could run into resistance from the South Korean government, which treats those technologies as sensitive, Reuters reported.

Elon Musk once again teased the potential of a merger between Tesla and SpaceX on Monday, saying it was a “great question” why Tesla and SpaceX were separate companies.

“With all this collaboration, on so many levels, who can imagine what action one might take when there’s so much close collaboration in so many areas,” Musk added, during a virtual appearance at the All-In Summit.

The question came up in reference to Tesla’s upcoming event for its next-generation Roadster. The vehicle was developed as part of a collaboration between Tesla and SpaceX, which built cold gas thrusters for a limited edition version of the hypercar, The Information previously reported. The event is scheduled to occur in Texas, near SpaceX’s rocket-testing facility, on Oct. 1.

During SpaceX’s first earnings call as a public company last month, Musk didn’t dismiss the possibility and said it would have to be “done with the appropriate process.”

Microsoft said Tuesday it will raise its quarterly dividend to 98 cents per share, up from 91 cents, beginning in December. The company previously raised its dividend by 10% a year ago. Microsoft’s stock was roughly flat in after hours trading.

The company’s move to return more cash to shareholders comes as it simultaneously balances growing spending on AI, with plans to spend $50 billion on capex this quarter alone, primarily to build new data centers to meet AI demand.

Still, Microsoft has also reported returns from its AI investments; it disclosed in July that it exceeded $100 billion in revenue from rentals of Azure cloud servers in the twelve months prior, and that it has more than 30 million paid subscriptions to its Copilot AI features for Office 365 applications, up from 20 million paying users at the end of the first quarter. Microsoft generated $20 billion in free cash flow in the June quarter, up more than $4 billion from the cash it generated in the first quarter but down 20% compared to its free cash flow in the second quarter last year.

DoorDash is teaming up with Marc Lore’s restaurant meal business Wonder, putting $125 million into the company and buying a campus dining business from Wonder for $300 million, the two companies said.

Wonder offers its own lineup of meals, made in its own kitchens, for delivery, which makes it something of a competitor to DoorDash, which delivers meals from a wide variety of restaurants. The deal raises the possibility that DoorDash could take over Wonder deliveries in the future.

The campus dining business, which Grubhub bought in 2018, lets college students order meals for delivery from campus restaurants using a mobile app. DoorDash plans to expand the business. DoorDash’s investment in Wonder will add to Wonder’s $650 million series D round announced in July, the companies said.

SpaceX CEO Elon Musk proposed that competing AI companies, including his own SpaceXAI, should “peer-review” each other’s models prior to release, as industry leaders acknowledge rising worries around AI safety.

“Instead of grading your own homework, you would at least have competitors grading your homework and raising the alarm if they see concerns,” Musk said at the All-In Summit in Los Angeles on Monday.

He said the proposal shouldn’t necessarily replace regulation, but could be more quickly implemented, and suggested that Chinese model makers could get on board.

Musk’s comments came in the wake of a series of hacking incidents involving AI and public warnings by AI employees that have highlighted the risks of the technology. In response, Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have expressed a willingness to pace the rate of AI development to improve monitoring and aligning models to human values.

Meta Platforms chief Mark Zuckerberg said ensuring the safety of AI is a commercial necessity for the companies that make models, and took an implicit swipe at calls for a collective slowdown in the technology’s development.

“Every lab has the responsibility and incentive to move at the pace required to train its models safely, and the ability to take its own actions to ensure that happens,” the CEO wrote in a post on X Tuesday.

Zuckerberg had been noticeably silent in recent days as other tech chiefs, led by Anthropic CEO Dario Amodei in an essay published Saturday, have warned that AI’s growing risks require new measures to limit advances at the technology’s frontier. Meta’s AI in recent years hasn’t been at that frontier in terms of performance and capabilities, though its most recent models have closed the gap considerably.

In his post, Zuckerberg argued that companies don’t need to wait for others to act, pointing to Meta’s delay in the release of its recently launched Muse consumer AI agent to work on safety and security issues. “We didn’t call for everyone else to do this before we would,” he wrote.

“There is a lot of debate about slowing progress on capabilities until alignment catches up,” he wrote, arguing that trust and alignment are becoming important competitive distinctions for AI models and agents. “Any lab that doesn’t focus on alignment will fall behind.”

At the opening of its annual Dreamforce customer conference, Salesforce signaled its involvement in two major trends in enterprise AI: Open-source AI models and software that helps corporate customers safely run AI agents on their data.

Salesforce unveiled a new AI model, called Koa, which it developed with Nvidia and is based on one of the chip maker’s Nemotron open-source models. Koa, a reasoning model designed using synthetic data, was customized in post-training to handle sales management tasks, from finding and vetting sales leads to dealing with customer service issues.

The other big reveal was AIForce, a new type of software for customers that use agents to access Salesforce data from outside its own applications—such as Anthropic’s Claude chatbot and Slack. AIforce ensures that agents can only access data they’re authorized to work with, while also giving them more precise definitions of data, such as the difference between gross revenue and net revenue. This helps agents complete their work faster and with better accuracy.

While Salesforce hasn’t gained much traction with its in-house AI models, its Dreamforce announcements demonstrate its ability, honed over more than two decades, to address the roadblocks that businesses sometimes encounter when adopting new technology.