WSJ What’s News - OpenAI Files for IPO in Test of Investor Appetite

Episode Date: June 9, 2026

A.M. Edition for June 9. OpenAI has privately filed for an IPO, setting the ChatGPT creator up to potentially listing as soon as this fall. WSJ tech reporter Sam Schechner says the filing comes amid intense competition with rival Anthropic and Elon Musk’s SpaceX and who will get the biggest slice of public investor money this year. Plus, the Pentagon targets Alibaba, Baidu and BYD in a new Chinese military blacklist. And from London Tech Week, our conversation with the founder of AI voice company ElevenLabs, Mati Staniszewski. Luke Vargas hosts. Sign up for the WSJ’s free What’s News newsletter. Learn more about your ad choices. Visit megaphone.fm/adchoices

Open AI files to go public and it'll be a major test of investor confidence that it can meet its computing commitments through 2030. Raising cash from public markets can help it finance those commitments. At the same time, it means that investors are going to get a lot more detail and are going to be able to scrutinize whether its revenue is growing fast enough to make good on those commitments. So this is both a super important way for them to raise capital, but it could be a double-edged store. Plus, the Pentagon adds Alibaba and Evie Maker BYD to a list of firms that it says are helping China's military and main Democrats stand by Senate candidate Graham Platner despite allegations of misconduct. It's Tuesday, June 9th. I'm Luke Vargas for the Wall Street Journal, and here is the AM edition of What's News, the top headlines and business stories moving your world today. OpenAI has confidentially filed IPO paperwork with the SEC, setting up the chat GPT creator to potentially go public as soon as this fall. The move comes after rivals Anthropic and SpaceX announced plans for mammoth public listings.

And while we don't have many specifics, OpenAI did say there's a complicated set of tradeoffs to going public, as tech reporter Sam Shackner explains. What we do know is that they are in an intense competitive rate. Anthropic recently surpassed OpenAI's private valuation. They are pulling ahead with business customers. OpenAI still has the bigger chat GPT consumer business, but it's a really tight race. And banks have told these two companies that whoever is first to the public markets could get the lion's share of the money from those markets. The company has hundreds of billions of dollars in commitments that they've made to building and

financing computing. And if they're going to finance those, they're going to have to find. the way to either generate that cash, to raise it through equity, or to borrow it in debt. Openangai recently completed the largest funding round in the history of Silicon Valley, raising $122 billion from Amazon.com, Nvidia, SoftBank, and other investors. The Pentagon has updated its annual list of Chinese military-linked companies, adding around two dozen new entities, including tech giants, Alibaba, Baidu, and electric carmaker BYD.

The designations limit the company's operations in America and bar them from doing business with the U.S. military, suggesting that national security officials see critical risks in Chinese commercial sectors far beyond just semiconductors and AI. Alibaba, Baidu, and other newly added firms have denied any ties to Beijing's military and are weighing actions to be removed from the list. Meanwhile, China's export engine is revving up with outbound shipments jumping nearly 20% in May, beating economists' forecasts. Chinese semiconductor exports more than doubled last month, thanks to the global AI boom, while shipments to the U.S. rocketed over 35 percent. The export surge is keeping Beijing on track to hit its 2026 growth targets, but is bound to reignite tensions with global trading partners who argue that China's state subsidies create an unfair playing field.

It's primary day in Maine, the site of one of the country's most competitive Senate races this year, In an incumbent Republican Susan Collins is set to face the likely winner of today's Democratic primary, oyster farmer, marine veteran, and political newcomer Graham Platner. That is, in spite of Platner being the subject of a number of controversies from his having a Nazi-linked tattoo to new reports from the journal that he engaged in sexually explicit texts with other women while married. National politics reporter Eliza Collins attended a town hall on Sunday where voters could ask Platner about allegations of past misconduct. conduct, but she told us that no one did. His supporters really stuck with him. There were zero questions about any of the allegations. Voters passed around a poster board where they wrote messages, including someone writing, everybody has a past, keep going, another person wrote, Stay Strong. For these supporters, I spoke to, these allegations really have not swayed them.

They are totally bought into Platner's saying that he does have a dark past. He has said he was a partner, that he was angry, but he has sought help. And for a lot of people I talk to, him basically owning up to these allegations was enough. Now, I should say that other allegations, including one of abuse from an ex, Plattner has denied. And we'll have results from that primary as well as other contests in Nevada, South Carolina, and North Dakota tomorrow morning. And Republican Spencer Pratt has failed to advance to a runoff election in L.A.'s mayoral race. That's according to the AP, which has Pratt in third place behind incumbent mayor Karen Bass and fellow Democrat Nithia Rahman with 93% of votes counted. Only the top two finishers will head to November's runoff.

President Trump has seized on California's laborious vote counting process in which officials count every ballot postmarked by Election Day and received within the following week in order to claim election fraud, despite there being no evidence of impropriety. Coming up, Meta launches a program to train blue-collar data center builders as it doubles down on AI investments. And I'll speak to the co-founder of conversational AI company 11 Labs about the potential of voice tech. That and more after the break. Meta is launching what it's calling a workforce academy to train skilled workers and land graduates with a job building the company's data centers. Paid training, a job, and a path. to America's future. The future is for everyone.

The program is set to launch in Louisiana, Ohio, Indiana, and Texas, and comes a little over a month after META laid off 8,000 white-collar workers in part to fund its AI infrastructure buildout. And so much for the death of the office, at least in Manhattan. The market there is on pace for its best leasing performance since the year 2000 when dot-com startups were fueling demand. And this time around, it's AI providing the tailwind. Through the first quarter, AI firms have now leased a million square feet of Manhattan office space,

already surpassing last year's totals with big deals signed in Soho and nearby Hudson Square, where we report that Anthropic is nearing a deal to lease a half a million square feet. AI is rapidly reshaping how the world thinks about voices. Our ticks, stutters, and twangs used to set us apart from the voices of machines, but now they can be replicated with remarkable accuracy for personal enjoyment or commercial gain. And helping to lead that charge is UK-based conversational AI company 11 Labs, which was valued this year at $11 billion. At this week's bustling London Tech Week, I caught up with co-founder Maddie Staniszevsky

and mentioned a recent experience speaking to impressively real estate, AI voices when calling a number of financial firms. They were really good. They got me the help I needed very quickly, but there was no, I mean, I was savvy to it. I realized this wasn't a real human, but there's no disclosures around that. Do you think that will become a trend, even if the quality of the voice is more convincing than it's ever been? So frequently when we speak with customers, we think it's the best is to preempt and lead

with that disclosure. Different countries will have different rules on whether you need to or not. But the second part here is, and recently we, we, we, we, we think, we, we, we think, we, we and us that work with Google and few other organizations on watermarking all the content that's generated. We've done it from the start. Now we are bringing CINFID as one of the common standards. We think this will be a common pace. All the content out there, yes, we'll need to be watermarked. And in some way, it will actually flip in the future. Now you are trying to detect is it AI or not. And the assumption is it's human. I think it will flip in the future.

It will be assumed to be AI and you'll detect for human. Everything else will be assumed to be AI. I mean, maybe some listeners will have done this before or know these examples. Like if you upload something on YouTube and it has music in it that belongs to a record label, it gets taken down immediately because they've uploaded their catalog in and it can spot the waveform exactly. I mean, is there a way to fingerprint a voice like that? It gets tricky, but yes, it's possible. You can watermark.

We found a way to embed it in audio without disturbing the audio itself to know that this belongs to A, Y, Z. Ultimately, yes. We think you can create a voice that belongs to a person and is uniquely theirs. We work with a lot of people with ALS and M&D, but lost their voices, and we work with them to bring them back. We partnered now for over 10,000 people, but recently we worked with an artist, a musician here in UK and Patrick Garling, and he used to perform in the band. And unfortunately, lost the voice, couldn't do that again, but we partnered again with him and his band,

so we could not only create the voice, we could create some version of a singing, and he could go back on the stage, and now is touring and performing with his band once again. and that iconic experience that you had and the way he performed what could be replicated and otherwise just not possible. How did you get there? Whose voices did you use to get this technology off the ground? So we assembled a speech from all types of dialects, languages, voices across for the model.

So usual techniques similar to other AI companies. And on top of that, because the voice that you use is slightly different to LLLM models, you need to choose the voice. And the way we've done that is we created a voice marketplace effectively, where anybody can create their voice, we authenticated, you can share it, and you earn compensation as a result. So we took an approach that we want to bring part of the industry

together with us. Today, we have 20,000 voices in the ecosystem. We paid back over $20 million back to that ecosystem. And that's all types of voice talent, people that have great voices that just wanted to be able to do that. Famous people, people that are coming. So a wide plethora of voices. I asked because there are some broadcasters,

you of course, Norther's case in Illinois, who are suing the company saying that their voices were exploited to train these models. I mean, was any of that done in the early iterations of getting this library built? Of course, I cannot comment on the ongoing matter. But we, you know, we will, we will, when I have more updates, I will be able to share them with you. But we feel we took and did something unique to the industry, as you probably seen, that same group is, I think, steering few other companies as well.

The reason I ask is I think they're a cynical takeaway of the AI industry as a whole, not speaking just about 11 labs, is that in the sort of first phase of development, you sort of maybe do stuff that you can't sustainably do in the future. You scale up, and if you can scale up, then you can change your practices in the future. I mean, am I being too cynical that that is something that has characterized how LLMs in particular were built?

Some companies definitely will take that approach. We are taking the approach where we want to lead responsibly from the beginning. But many parts which is exciting for us is, and a thing that's maybe in contrast to some of the other companies, is like, how can AI and the industry work together? Yesterday, the UK government signed an MOU with 11 labs to explore using its tech to make government information and services more accessible, including via multilingual support.

And that's it for what's news for this Tuesday morning. Today's show was produced by Hattie Moyer. Our supervising producer is Sandra Kilhoff, and I'm Luke Vargas for The Wall Street Journal. We will be back tonight with a new show, and until then, thanks for listening.