MARKETS

Brutal hype test for AI IPOs arrives with SpaceX

The summer of AI IPOs has arrived, and it's off to a strong start. SpaceX closed up 19% at roughly $161 per share on Friday, priced at $135 at IPO, delivering the largest market debut in history at a $1.77 trillion offering valuation, which climbed above $2 trillion by close.

While best known for its low-orbit satellite network via Starlink, SpaceX also has a foothold in the AI world through its $250 billion acquisition of another one of founder Elon Musk's startups, xAI, which includes the Grok chatbot and the Colossus gigawatt-scale AI factory. With OpenAI and Anthropic filing confidentially for their own IPOs earlier this month, the question is: Does SpaceX's IPO success set the tone for how public markets will value Anthropic and OpenAI?

Here's why those scenarios are significantly different and difficult to compare: SpaceX has a highly profitable Starlink business with a 39% operating margin, which can offset any losses from the AI arm. Those losses are largely unavoidable due to data center supply constraints that drive up training and inference costs. When factoring in the scale of OpenAI and Anthropic's free services, which are needed to secure users, they are bleeding money. So much so that, ahead of each company's IPO, they are considering substantial token price cuts to capture more customers within their respective customer segments amid consistently rising costs.

Another major difference is SpaceX's promise to solve the aforementioned data center crisis. The $250 billion xAI acquisition was in part justified because of plans to build space-based data centers, which aim to overcome the limitations of terrestrial AI infrastructure — plans that are now moving markets. 

As noted by Morningstar in a research note, SpaceX's valuation would be considerably lower than its IPO valuation (the firm set it at $63 per share) if not for some of its more ambitious future projects, including orbital data centers. Yet markets are clearly excited about the AI arm.

"The [higher IPO] valuation hinges on two unproven technologies: a rapidly reusable Starship upper stage and commercially scalable and competitive orbital AI data centers. We expect neither of these technological questions to be answered before 2028, even in the most optimistic scenario," said MorningStar in the note. 

Furthermore, the firm adds that in the most ambitious moonshot scenario, SpaceX could rapidly scale orbital data centers to capture 20% of the firm's forecasted AI computing capacity by 2040, implying a share price of $154 per share, but it assigns that scenario only a 7% probability.

SpaceX's success may not be a perfect blueprint for Anthropic or OpenAI, but that doesn't preclude both companies from eventually going public at similarly eye-watering valuations. Demand for AI products remains as strong as ever, and the capital flowing into the sector shows no sign of slowing. Anthropic and OpenAI are currently valued at $965 billion and $852 billion, respectively. Those figures, combined with SpaceX's valuation, put the total ask on public markets well above what the entire US IPO market raised in 2025. That scale alone will inevitably reignite debate about whether we're watching a bubble form in real time.

TOGETHER WITH IBM

Rethinking how developers work with AI

Vibe coding brought generative AI into the flow of development fast—easy to prompt, quick to iterate and useful for early exploration. Turning that momentum into production-ready software calls for more discipline.

Agentic engineering treats AI agents as part of a managed workflow, where developers guide decisions, review outputs and stay accountable for results. It supports everyday tasks such as refactoring, testing and documentation, while adding structure, traceability and governance. 

Teams can gain a clear path from experiment to deployment, with control. This approach helps developers scale AI adoption while supporting quality, consistency and long-term maintainability across modern software systems. It aligns AI use with enterprise standards, enabling teams to integrate, extend and operate systems in production environments.

Discover what’s next

BIG TECH

Google’s AI scam fight may be coming too late

AI has supercharged scammers with an entirely new arsenal of tools. Google is cracking down. 

On Friday, the company filed a lawsuit to dismantle a Chinese cybercrime group known as "Outsider Enterprise," which it alleges has used Gemini and other AI tools to generate thousands of fake text campaigns. According to Google, the group has used AI-generated government and brand websites to obtain people's credit card numbers and personal information, using the targets' trademarks and logos in these phishing campaigns. 

"You’ve seen the texts: fake package alerts, urgent bank warnings, panicked messages about your compromised account," the company said in its announcement. "Behind them is an AI-powered cybercrime network built to steal your passwords and credit cards."

The tech giant claims that more than 100,000 Americans have been impacted, and losses are "estimated in the millions." Additionally, more than 9,000 fake websites and one million fraudulent URLs are connected to this group. 

But the lawsuit isn't the only action Google is taking against these scams: 

• The company said it is coordinating with the FBI to go after Outsider Enterprise, as well as working with telecom providers like AT&T, T-Mobile and Verizon to block texts before they reach users. Halimah DeLaine Prado, Google's general counsel, told The New York Times that it represents the company's first coordinated effort and lawsuit of this nature. "That speaks to the breadth of impact that this particular scam has," she said.  

• The company is also backing seven bipartisan bills that aim to fight back against these scams, including legislation that specifically protects elders against these kinds of attacks, and a bill for public awareness and education around AI. 

It's also not Google's first attempt at warding off scammers: In early June, Google launched fake call detection for Android, which flagged suspected spoofed calls. According to Google, 55,000 spam texts were flagged by Android users over a two-week period in May, and more than 2.5 million messages by this organization were sent to Android users during that same period.

AI has given both security defenders and attackers way more ammo. Through this effort, Google is taking accountability for its part in that, tracking down the wide-scale misuse of its models. Though Google's attempt to contain the problem is a noble one, and some of the damage could be mitigated, it's still a reactionary effort rather than a proactive one. It begs the question of whether the power that these models provide should be so widely available for anyone to use or abuse. Additionally, if the genie is already out of the bottle, should the industry heed the slowdown warnings from companies like Anthropic before more harm is done?

IN PARTNERSHIP WITH LAMBDA

Cut your AI training costs by 25% or more

Most large-scale AI training runs use less than half the computing power they're paying for. Lambda's team found the root causes and built a reproducible framework that boosted efficiency by over 25%, without changing the model itself.

Lambda’s whitepaper shows you how to address:

• Memory inefficiencies silently inflating your costs

• Training configurations that aren't making full use of your hardware

• Bottlenecks that slow down GPU communication

Get the whitepaper.

WORKFORCE

The 3 human skills that will endure in the AI era

Perplexity has become one of the most important AI companies in the world, but its ambitions now stretch far beyond AI-powered search.

In this episode of The Deep View Conversations, we sit down with Dmitry Shevelenko, chief business officer at Perplexity, to discuss how the company evolved from an AI answer engine into a platform for AI agents and digital coworkers. Dmitry explains why Perplexity has focused so intensely on accuracy, how AI is changing the nature of work, and why he believes the future belongs to small, highly leveraged teams.

The conversation also explores Perplexity Computer, hybrid compute, the coming shift toward AI agents, and what leaders need to do to stay relevant in a world where AI increasingly performs knowledge work.

Topics covered:

• Why Perplexity made accuracy its defining principle

• How Perplexity grew from 20 employees to 400

• The rise of AI agents and digital coworkers

• Why Perplexity abandoned advertising as a core strategy

• How Perplexity Computer orchestrates multiple AI models

• The future of hybrid cloud and local AI computing

• Why "tokenmaxxing" may not be sustainable

• How AI is reshaping entry-level jobs

• Why entrepreneurship may become the new career path

• The three skills that will matter most in the AI era

• How leaders should think about leverage and productivity

• What Perplexity sees coming next in AI

If you're trying to understand where AI agents are headed, how work is changing, and why Perplexity has emerged as one of the AI industry's key players, this is a conversation you won't want to miss.

📺 Watch on YouTube
🎧 Listen in your favorite podcast player

Subscribe to Deep View Conversations for interviews with the leaders shaping the future of AI, business, and technology.

LINKS

• US govt moves to impose export controls on Anthropic's latest models

• State attorneys general combine to launch lawsuit against OpenAI

• KPMG has pulled a report on AI agents due to reported hallucinations

• ChatGPT has eclipsed a billion monthly users

• Zuckerberg admits Meta made mistakes in its organizational shift around AI

• French AI startup Mistral in talks to raise €3 billion at a valuation of roughly €20 billion

• Perplexity Plan Mode: This feature clarifies unclear questions with users before completing tasks. 

• Rebel Audio: The AI-powered podcast studio has launched in public beta. 

• ChatGPT Model Picker: OpenAI has updated its chatbot's model selection tool with an updated list of options.

• Project Genie: Google has expanded access to its world model project to AI Ultra 5X subscribers.

• Sesame: Research Scientist

• Meta: Partner Engineer, Generative AI

• Thinking Machines: Research, Pre-Training Science

• Anthropic: Research Engineer, Knowledge Foundations

GAMES

POLL RESULTS

Have you or a colleague ever used unauthorized "shadow AI" tools at work?

Yes (29%)
No (47%)
I’m not sure (14%)
Other (10%)