Starship V3 Scrub Delays SpaceX’s IPO-Timed Reuse Test

SpaceX’s 12th Starship test flight was halted at roughly T-minus 40 seconds, but the significance of the scrub was larger than the part that failed. The attempt was set to debut Starship Version 3, a redesigned vehicle meant to move the program toward much higher payload capacity and full reuse, shortly after SpaceX had formally filed for an IPO, as Ed Ludlow framed it.

Loren Grush said Elon Musk described the immediate problem as a hydraulic pin controlling the launch-tower arm. That arm, along with related quick-disconnect systems, has to move rapidly out of the rocket’s path once launch begins. A small mechanism, in other words, was positioned at a critical point in the sequence.

Grush said SpaceX also appeared to be working through other problems, including an issue involving the flame trench that had stopped the countdown earlier. She described the caution as understandable: the launch involved a new pad, a new rocket iteration, and a test that SpaceX would particularly want to go well given the IPO filing. SpaceX, she said, may have been “a little extra cautious” compared with prior Starship test missions.

A Bloomberg graphic displayed Musk’s post explaining the abort: “The hydraulic pin holding the tower arm in place did not retract. If that can be fixed tonight, there will be another launch attempt tomorrow at 5:30 CT.”

The larger test was Starship V3. Grush described the new version as the vehicle intended to achieve the ambitions SpaceX has laid out for Starship: roughly 100 metric tons of payload capacity and optimization for total reuse. Ludlow contrasted that with prior iterations associated with about 35 metric tons of payload to low Earth orbit.

Grush called full reuse the program’s “ultimate holy grail”: bringing back both Starship and the Super Heavy booster intact so they can be flown again quickly. That is why the abort over a pin did not land as a trivial delay. It interrupted a test of the version SpaceX expects to move Starship closer to its stated performance and reuse goals.

Laura Crabtree, CEO of Epsilon3 and a former SpaceX senior mission operations engineer, said Starship V3 matters because the industry has been waiting for an operational version that can carry larger payloads more affordably. In her description, Starship is an enabling platform for larger cargo, lunar rovers, more people going to the Moon, and easier access to space.

Ludlow connected that capacity to SpaceX’s IPO filing, saying larger future payloads — including orbital data centers — were a main focus of the filing. Crabtree widened the frame to SpaceX’s founding ambition of making humans a multi-planetary species. Starship, she said, is a “next piece of the puzzle” alongside Starlink, data centers, and other infrastructure SpaceX has built toward that goal.

Crabtree stopped short of describing crewed Starship as a finished product. Asked where astronauts would physically go in a Starship configuration, she said there had been many design iterations and that she was not entirely sure where the design stood. The last version she had seen treated Starship itself as the human-oriented payload vehicle for lunar transport, with toilets and pods for people to live in during the trip. But she said the public likely would not know the exact design until Starship is operational and has flown many times, and suggested SpaceX was probably waiting to unveil those details in later versions.

Her caution mattered because the claims around Starship depend on repeated proof. The vehicle is being discussed as an architecture for lunar and eventually interplanetary transport, but its crewed design, operational cadence, and reuse profile still require testing and disclosure.

Crabtree also described the internal engineering culture behind SpaceX’s pace. Engineers, she said, are given problems that can feel “somewhat insurmountable,” more responsibility than they might receive in a traditional aerospace company, and the expectation that they will solve, test, and iterate quickly. That helps explain the fast iteration in Starship, she argued, because “you can’t design something like Starship with the things that have been done in the past.”

Asked how difficult the current Starship program is, Crabtree described a system with “millions of decisions” and “millions of technical things” that must go right. In that context, a scrub caused by a pin and possibly a few out-of-family temperatures was not, in her view, an indictment of the program. Scrubs and some failures should be expected. The goal is to test enough small things before the final mission to have confidence that the mission will go well.

Lenovo shares rose nearly 20% in Hong Kong trading after the company reported strong AI-related growth, a move Ludlow described as the biggest single-day jump since 2008. A Bloomberg market graphic listed Lenovo at 15.75 Hong Kong dollars, up 2.60, or 19.77%, intraday.

Winston Cheng used the earnings reaction to argue that Lenovo should not be understood only as the world’s leading PC maker. He said Lenovo has a full device portfolio across PCs, tablets, and smartphones at global scale on a non-iOS basis, and that this device base is the foundation for AI devices. He also pointed to Lenovo’s “pocket to cloud” hybrid strategy, with infrastructure used for AI training and inferencing.

Cheng described the current period as the “AI decade” for Lenovo, but said the company was only in the first year of that journey. He said Lenovo had invested in relevant capabilities for more than 10 years, and that its ability to produce at scale globally on both the device and infrastructure sides is distinctive. He also cited Lenovo’s IBM heritage from the 2015 acquisition as part of the foundation for serving AI compute needs, especially through high-performance computing and liquid cooling.

A Bloomberg earnings graphic listed revenue of $21.59 billion, up 27% year over year; R&D expenses of $747.9 million, up 16%; and gross margin of 16.4%, unchanged year over year.

Asked for evidence that AI agents and inference are driving demand for new enterprise PCs, Cheng said Lenovo’s AI PC demand was “very strong” and increasing as a share of the portfolio. He said AI revenue represented about 38% of Lenovo’s total revenue in the latest quarter, with AI PCs a significant component. His argument was that as AI training develops and AI capabilities improve, more people will consume and interact with agents through devices, creating demand for higher-performance hardware.

Cheng also pointed to broader market signals: AI companies, including OpenAI and others, are pursuing partnerships with device makers either through staffing or strategy, and Lenovo is seeing similar dialogue with AI companies that want to work with it on go-to-market.

The constraint is components. Cheng said AI infrastructure spending is driving shortages not only in memory but also in CPUs and GPUs. He described the demand as a multi-year opportunity but said the supply chain still needs to catch up. For now, the imbalance is pushing memory prices higher.

On AI infrastructure, Cheng said Lenovo can pass through those pricing pressures because demand is strong. The device side is more complicated, especially at the lower end of the market, where he said companies that cannot get supply have seen market-share declines. Asked whether Lenovo has to choose between allocating scarce components to servers or PCs, Cheng said suppliers value Lenovo’s complete global portfolio and its long-term forecasts across PCs, tablets, smartphones, and servers. He said Lenovo still needs enough supply for longstanding ThinkPad demand.

Oura has confidentially filed for a U.S. public listing, according to Bloomberg reporting cited by Ludlow, with the company later confirming the filing. A Bloomberg News graphic said the number of shares and price had not yet been determined; Goldman Sachs, Morgan Stanley, and JPMorgan were working on the filing; and the company aimed to go public later in the year.

Tom Hale had recently offered a version of the investor pitch Oura may use: hardware is attractive because “you can’t vibe code atoms.” Hale said Oura’s identity as a hardware company with strong software backing had become an advantage, because investors were glad it was not merely a software company.

Dana Wollman said the confidential IPO filing put an exclamation point on the mainstreaming of smart rings. She described Oura as having moved from a niche hardware category — one journalists long had to explain — toward a product category that may become widely understood. Sales have been rising, she said, and the company may be broadening beyond an audience that has historically skewed female.

Wollman also distinguished Oura from smartwatches such as the Apple Watch. In her account of Hale’s view, Oura is not trying to be another wrist screen with notifications. It is trying to occupy its own lane: a health and fitness tracker for people who want less distraction, with the longer-term ambition of becoming a kind of auxiliary predictor of overall health, not just a step counter.

Oura’s case is a different version of the hardware argument from Lenovo’s. Lenovo is describing AI hardware at global device-and-infrastructure scale. Oura’s pitch, as framed by Hale and Wollman, is that specialized consumer hardware tied to health data and software can be compelling precisely because it is not easy to reproduce with code alone.

Salesforce has been making prominent claims about agentic AI capabilities, but Bloomberg’s reporting found that some showcased functionality is aspirational rather than broadly usable today. Brody Ford said Bloomberg applied a simple test to one of tech’s most visible marketers: when Salesforce shows AI agents handling customer service or other work by themselves, can customers actually use that functionality now?

In many cases Bloomberg tried, Ford said, “it’s not there yet.” The commercials and keynote demos show a future vision that customers hope to implement. His broader point was not that Salesforce is uniquely guilty of overselling, but that getting AI systems online inside real companies is difficult and will be “a bit of a slog” in many cases.

The practical distinction is between signing up for an agentic product and deploying it. Ludlow raised that issue directly: Salesforce may cite the number of customers signed up for an agentic product, but the relevant question is whether those customers are using it in the real world. Ford said he could not find hard data on that. He offered Williams-Sonoma as an anecdotal example: the retailer had appeared in keynotes and commercials, but when he asked about some of the more advanced functionality shown on stage, he was told the hope was to have it up by the holidays.

Ford said Salesforce CEO Mark Benioff defended the marketing as normal for the tech industry. In Benioff’s view, companies have to show customers a vision; everyone does it; customers are not confused; and the disclosures are proper. Ford was less certain that ordinary viewers of a commercial would understand that a demonstrated feature may not yet function in the way shown.

That exchange sharpened one of the day’s recurring AI tensions. Executives at Lenovo and Zoom pointed to revenue, usage, and customer demand as evidence that AI is already part of their businesses. Salesforce, as reported by Ford, illustrated a different risk: agentic AI can become part of a company’s commercial story before the most ambitious versions shown on stage are deployed.

Zoom’s stock rose after the company raised its full-year forecast for adjusted earnings and revenue. Bloomberg graphics showed Zoom trading at about $108 intraday, up more than 11%, and listed analyst moves: KeyBanc raised the stock to sector weight from underweight, RBC lifted its price target to $130 from $110, and Baird lifted its target to $115 from $105.

Michelle Chang said the quarter reflected work Zoom had been doing consistently over several periods: becoming “no longer just the meetings company.” She argued that the results showed AI monetization, an inflection in growth, and continued profitability.

Zoom’s competitive problem is familiar: many users are invited to meetings on other platforms, including tools from larger technology companies. Chang said that a few years ago Zoom’s pitch centered on quality, reliability, and keeping meetings from failing. She said the company is now positioning itself around a “system of action” — using AI across what Zoom sees inside an organization and outside it, including customer experience.

Her claim was that Zoom is moving toward “real tangible AI value” already delivered to customers. Ludlow pointed to paid users of Zoom’s AI Companion growing sharply. Chang said paid monthly active users rose 184%, and that AI use began with meeting summaries. She highlighted a newer feature, My Notes, described as a personal context and note-taking tool, which she said grew to 1.5 million monthly active users in four months.

Chang’s broader description of Zoom’s direction was the “meetings lifecycle.” The value proposition, she said, is no longer confined to the 30 or 60 minutes of human-to-human connection, which she argued will endure in an AI world. Instead, Zoom wants to help with the full lifecycle of getting work done: taking conversations and moving them to completion.

She also tied that positioning back to financial performance: profitability, revenue growth, and high cash-flow generation. Zoom’s AI story, as Chang presented it, rests on paid usage and workflow expansion rather than only on a future-agent promise.

Several shorter items extended the same set of business questions without carrying the weight of the main interviews.

Waymo temporarily halted service in five cities during severe weather because of concerns that its robotaxis might attempt to drive on flooded roads. Ludlow said storms in Atlanta included an unoccupied Waymo vehicle driving into and getting stuck on a flooded road.

DeepSeek senior management, according to sources cited by Bloomberg, told potential investors that the startup would prioritize groundbreaking AI research over short-term monetization. The company was said to be in final discussions over a funding deal that could push its valuation to about $45 billion before the investment.

The European Union, according to sources cited by Bloomberg, planned to propose temporarily lifting sanctions on a major Chinese chip supplier after automakers warned of impending supply-chain chaos if the ban remained in place. Such a move would still require approval by the bloc’s 27 member states.

The clearest workplace-governance example came from Standard Chartered. CEO Bill Winters had described AI investment as replacing, in some cases, “lower-value human capital” with financial and investment capital. Ludlow said the phrase prompted public backlash. Winters later wrote on LinkedIn that the bank has long invested in helping staff whose roles are upended by automation and said it had a responsibility to help colleagues move into higher-value roles. A subsequent LinkedIn post apologized for the wording, saying his choice of words had caused upset to some colleagues: “And for that I am sorry.”