← Tech Brew Ride Home
Tech Brew Ride Home
Meta Needs To Rethink Manus
00:21:17
China blocked Meta's $2B Manus acquisition and ordered both sides to unwind the deal, closing the "Singapore washing" loophole for Chinese AI startups. OpenAI is developing smartphone chips with Qualcomm and MediaTek, Google controls ~25% of global AI compute, and SaaS pricing shifts to usage-based.
Loading summary
Transcript6 lines
- [00:04]
A
Welcome to the Tech Brew Ride home for Monday, April 27, 2026. I'm Brian McCullough. Today, China blocked Meta's $2 billion Manus acquisition and ordered both sides to unwind the deal, closing the so called Singapore washing loophole for Chinese AI startups. OpenAI is developing smartphone chips with Qualcomm and MediaTek. Google controls around 25% of global AI, compute and SaaS. Pricing shifts to usage based here's here's what you missed today. In the world of tech, If there's one thing that slows your flow state to a crawl, it's switching between 15 tabs. Microsoft Azure brings all of your essential tools together in one unified experience. You can build, test and deploy without switching platforms or losing Steam. Azure delivers enterprise grade security, compliance and governance to help give IT admins and developers the confidence to ship faster. And whether it stems from a viral moment or a a global expansion, Azure can scale automatically when you need it, so success doesn't turn into downtime or middle of the night firefighting. Try Azure for free for 30 days with more than 65 always free services@azure.com getstarted that's a z u r e.com get started China is blocking Meta's $2 billion Manus acquisition after reviewing whether it violated investment rules and is tell both parties to cancel the deal. Manus moved to Singapore in 2025, but apparently that didn't save them. Quoting the Times, Chinese officials had said in January they were investigating whether Meta's acquisition of Manus last December violated the country's rules on foreign investment. They were also assessing whether the deal violated China's requirement that companies obtain approval for the export of certain technologies. The National Development and Reform Commission, a high level ministry that oversees economic planning and plays a central role setting China's AI policy, said on Monday that it had decided to prohibit foreign investment in Manus and instructed the parties involved to withdraw the acquisition. It is not clear how such a transaction would be unwound, however. Meta has described the two teams as deeply integrated. Members of the Manus team have been working alongside Meta colleagues at the company's office in Singapore, according to two people familiar with the operation who spoke on condition of anonymity. Meta did not immediately respond to a request for comment. The company previously said that the transaction had fully complied with applicable law. The Chinese government issued its decision just a few weeks before a planned meeting between President Trump and China's leader, Xi Jinping. The New York Times reported last month that officials from the Chinese agency had called in Meta and Manus executives to express concerns about the deal and that Manus executives had been restricted from departing China as part of an apparent effort to discourage Chinese AI executives from moving businesses offshore. As companies in China and the United States race to develop cutting edge artificial intelligence, the scrutiny could make it harder for other Chinese firms to attract funding from foreign investors. It could also signal to Chinese researchers not to follow the path Manus took, in which Chinese executives register companies outside China to sidestep regulations from both Washington and Beijing. Manus is based in Singapore but was founded by Chinese engineers and had a Chinese parent company. The company was incorporated offshore and set up in China as a foreign owned entity. It has affiliated offices in Beijing and Wuhan. Many Chinese tech founders hope to attract Silicon Valley investors, but in recent years they have increasingly found themselves needing to choose between targeting the Chinese market or moving their headquarters outside China to court foreign investors. Jiang Gang Li, chief executive of Momentum Works, a consultancy in Singapore, said that scrutiny like the Manus deal is facing will, quote, make it increasingly hard for Chinese AI founders who started in China to sit on both sides or switch to the other side. There are already a lot of uncertainties starting an AI startup and most founders are technologists but not politically savvy, Mr. Lee said, end quote quoting Howdy MarionX. China just killed the $2 billion meta acquisition of Manus, and the implications go way beyond a single acquisition. Meta now has to unwind the deal. This decision closes the offshore arbitrage window that enabled Chinese founders access to Western venture capital and the ability to build Western facing products without Chinese domestic content restrictions. Manus relocated from China to Singapore, a practice known as S, to access deeper capital pools from foreign investors like Benchmark. The assumption made was location of the holding company equaled regulatory jurisdiction. Beijing's decision means that this offshore jurisdiction is not beyond their reach. Developing technology in China before transferring assets to an overseas entity through restructuring is now a red flag. Where you build your product matters more than where the holding company is registered. Some context Washington already banned Western investors from backing Chinese AI companies companies this is Beijing reciprocating. In a sense there is no more middle lane. End quote quoting Neil Shah on X this is going to be popcorn emoji. As I understand it, Manus is already deeply integrated into Meta's systems and Meta Superintelligence Labs will be interesting to see how this deal takes an alternative form if it doesn't go through. End quote and quoting Li Ling Y on X A reality check for the debate over Chinese investment in the U.S. beijing just showed how quickly they can shut the door on the reverse. The NDRC's blocking on the Meta Mana Steel is as clear a statement of intent as you'll get. Not surprising, but very telling. End quote. Haven't heard this guy's name in a while Our old friend Ming Chi Kuo says that OpenAI is working with MediaTek and Qualcomm to develop smartphone chips, with Luxshare handling the system co design. Mass production of these smartphone chips is expected in 2028. Quoting Bloomberg Qualcomm shares jumped in pre market trading on Monday after a closely watched tech industry analyst suggested the chip maker is working with artificial intelligence giant OpenAI on a smartphone. Shares of the San Diego based company gained 14% in early trading following a post on X by TF International. Securities analyst Ming Chi Kuo, which said his latest industry checks had revealed that OpenAI was working with MediaTek and Qualcomm to develop smartphone processors, added that Luxshare Precision Industry is working as the exclusive system co design and manufacturing partner. Luxshare shares jumped as much as 10% in Shenzhen trading. Meanwhile, shares of iPhone maker Apple fell as much as 1.9% in New York, the report says. Mass production is expected in 2028, although specifications and suppliers are expected to be finalized by late 2026 or the first quarter of 2027, Kuo said. Kuo's report comes at a time of uncertain prospects for Qualcomm, which has seen its shares struggle this year. While stock had jumped about 20% off a recent low through Friday's close, It remains down 13% in 2026, making it by far the worst performer in the Philadelphia Semiconductor Index, which has gained for a record 18 straight days and is up nearly 50% this year, end quote quoting Android Police Instead of raw power, OpenAI's smartphone system on a chip will likely focus on faster AI performance, something which Google already does with its Tensor chips. This also explains why it's collaborating with Qualcomm and MediaTek as it sa OpenAI from developing its own CPUs and GPUs from scratch, a time consuming and expensive process. OpenAI is already working with Broadcom on its own AI chips for data centers. Such custom chips could help speed up AI workloads and reduce costs. Google, Amazon and Meta are also building their own AI chips for similar reasons. Kuo predicts that AI agents will become a key feature in future smartphones. Instead of apps, users will rely on AI agents to get things done. This will require phones that feature on device and cloud AI integration. For OpenAI to deliver such an experience, it will need full control over the hardware and software. The analyst also believes that OpenAI may bundle subscriptions with the phone and work with developers on building an AI agent ecosystem. Essentially, like Apple, OpenAI will want to use the hardware to lure more customers to subscribe to its services. OpenAI is not the only company working on such a future AI first smartphone, Google, Apple and other companies are also possibly working on such initiatives for their future devices.
- [08:56]
B
Ready to soundtrack your summer with Red Bull Summer All Day Play, you choose a playlist that fits your summer vibe the best. Are you a festival fanatic, a deep end dj, a road dog, or a trail mixer? Just add a song to your chosen playlist and put your summer on track. Red Bull Summer All Day Play Red Bull gives you wings. Visit RedBull.com BrightSummerAhead to learn more. See you this summer.
- [09:22]
C
You can't reason with the sun. Trust us, we've tried this summer, it's time to put that angry ball of fire on mute. Columbia's Omnishade technology is engineered to protect you from the sun's harsh rays that can burn and damage your skin. The sun is relentless, but so is our gear. Level up your summer@columbia.com to spend more time outside and less time slathering on aloe lotion. You're welcome. Columbia engineered for whatever
- [09:52]
A
interesting one to pair with that one. The Financial Times says that Google controls around 25% of global AI compute right now, with around 3.8 million TPUs and 1.3 million GPUs. Google Cloud CEO Thomas Curran says demand and revenue justify the spend they've been making on all of this. Quote Thomas Kurian said that after a slow start in AI and entering the cloud business late, Google's full stack AI strategy, which includes building chips, data centers, foundation models and products in house, was starting to pay off. We're not just a hyperscaler reselling other people's technology. Our differentiation comes down to the fact that we own the ip, the model and the chips are ours, kurian said in an interview. For every dollar of revenue, we're not shipping 80% of it to either a model or chip provider, which allows us to invest more, he added. AI is now helping Google Cloud to grow faster than its rivals. It reported a 48% jump in revenue in the final quarter of 2025 and is on track to generate more than $70 billion this year, up from $43 billion in 2024. Google believes its TPUs and Gemini models are far ahead of Amazon's Trainium chips and Nova AI system, as well as Microsoft's Maya processors and MY models. This makes the search giant less dependent on partnerships with Anthropic or OpenAI or on Nvidia's expensive GPU chips. Kurian said that Google's 12 year investment in DeepMind allowed it to continually improve its proprietary chips and deliver consumer and enterprise AI products at a lower cost with better margins. A report from Epic AI estimates that Google controls about a quarter of global AI computing power, about 3.8 million TPUs and 1.3 million GPUs. Microsoft is second with 3.2 million Nvidia GPUs. In a recent podcast, Nvidia Chief Executive Jensen Huang criticized Google for not submitting its AI chips to independent tests and cast doubt on their performance and efficiency claims. He added that 100% of demand came from Anthropic alone and without the startup, why would there be any TPU growth at all? End Quote. Community Service Note Something you should be aware of the way you're going to pay for SaaS is changing. Quoting the information Dozens of enterprise software firms have shifted away from charging customers flat per user subscription fees as AI threatens their seat based pricing model. By the end of 2025, 79 of the 500 software companies tracked by former OpenView partner Kyle Poyar, including HubSpot, Adobe and Salesforce, had begun charging customers additional fees based on how much AI they're using. That's more than double the figure in 2024. The changes came after customers paying a flat fee for AI features and enterprise apps increased usage, raising costs for the app makers. The shift was also prompted by concerns that customers will require fewer subscriptions if they rely on AI agents rather than employees to interact with enterprise apps. Firms such as ServiceNow and Workday this year have been touting or ramping up usage based pricing, charging customers their AI tools in part based on how much they use them as measured in bits of data the AI processes. Another major firm, Atlassian, said it will soon charge such consumption based fees for its AI features, which customers use to search for files, draft documents and summarize meeting notes. Previously, many of the software providers sold AI tools as part of premium software subscriptions for core enterprise apps. Traditional enterprise firms aren't the only ones making the shift, however. Anthropic and OpenAI also revamped their pricing models in recent months to charge more for enterprise customers based on AI consumption. The change at Anthropic came after customers paying a flat fee increased usage of coding and other products, raising Anthropic's costs or hit usage limits of the subscription tiers. But consumption based pricing can lead to ballooning costs. Most of my clients hate it. The costs go through the roof really quickly, said Adrian Balfour, founder and chairman of AI Consultant and Vorso, about pricing models tied to usage. Even sophisticated technology firms such as Uber are having trouble adjusting to AI pricing changes, especially around coding products. Uber Chief Technology Officer Praveen Nepali Naga said the company blew through its full year AI budget in just a few months into 2026. End Quote. Finally today, from the Welcoming our AI Bot overlords file quoting the decoder In a week long experiment, Anthropic let CLAUDE agents buy and sell goods for employees. The result was stronger models negotiated better prices. The catch was the people stuck with weaker agents had no idea they were losing out. In December 2025, Anthropic ran a one week classifieds marketplace experiment called Project Deal for 69 employees at its San Francisco office. The whole thing ran through Slack, with CLAUDE agents handling every negotiation and deal. Each participant got a $100 budget. Before things kicked off, CLAUDE conducted a short interview with each volunteer to figure out what they wanted to sell at what price, what they to buy, and what kind of negotiating style their agent should use. Anthropic then turned those answers into a custom system prompt for each agent. From there, the agents took over completely. They wrote listings, found potential buyers and sellers, made offers, haggled over prices, and closed deals without checking in. The humans only stepped back in at the very end to swap the actual items, which ranged from a snowboard to a bag of ping pong balls. The real research question was hidden in a parallel experiment that participants didn't know about About. At First, Anthropic ran four versions of the marketplace at the same time. In two of them, every agent used Claude Opus 4.5 Anthropic's Frontier model at the time. In the other two, each participant had a 50% chance of being represented by Claude Haiku 4.5 instead Anthropic's smallest model. Either way, only the AI agents talked to each other. In the real run, where every agent used opus, the 69 agents closed 186 deals across more than 500 listings, moving just over $4,000 in total. Participants rated the fairness of individual deals at 4 out of 7 on average. Right in the middle, the mixed runs exposed a measurable gap. OPUS users came out ahead, closing about two more deals on average than Haiku users When the same item sold once through an Opus agent and once through a Haiku agent, opus pulled in $3.64 more on average. A Lab grown Ruby, for example, sold for $65 with Opus but only $35 with Haiku. The Opus agent opened at $60 up by competitive bidding, while the Haiku agent started at $40 and got talked down. Across 161 items sold in at least two of the four runs, an Opus seller pulled in $2.68 more on average, while an Opus buyer paid $2.45 less. When an Opus seller faced off against a Haiku buyer, the Average price hit $24.18 compared to $18.63 for Opus. On Opus deal, with a median price of $12 and an average of $20.05 across all runs, Anthropic says these gaps aren't trivial. The negotiation instructions participants gave their agents barely mattered. Some asked for a friendly approach. Others wanted aggressive tactics like negotiate hard and lowball. At first, aggressive sellers did get higher prices, but only because they set higher opening prices to begin with, Anthropic says. Despite the clear price gap, participants with Haiku agents rated the fairness of their deals almost the same as opus users for profit 4.06 versus 4.05 on the fairness scale. There was almost no statistical meaningful difference in satisfaction with individual deals. Of 28 participants who used both Opus and Haiku across different runs, 17 preferred their opus run, but 11 actually preferred the Haiku run. Anthropic calls this an uncomfortable implication. When agents of different strengths meet in real markets, people could end up on
- [18:08]
D
the the Wired newsroom is known for award winning reporting on how technology shapes our world. On WIRED's Uncanny Valley, we take that curiosity even further. Each week, journalists from Wired break down the biggest stories in tech while speaking directly with the people building challenging and reshaping the future. Is the AI boom sustainable? How do you protect your privacy in an age of constant surveillance? Uncanny Valley tackles the questions driving today's tech debates and lighting up your group chats. Listen to new episodes every Thursday wherever
- [18:39]
A
you get your podcasts losing side without ever knowing it. The company admits the experiment wasn't built to dig into these dynamics in detail and says more research is needed. The experiment also suggests that AI agent commerce isn't some far off scenario 46% of participants said they'd pay for a service like this. At the same time, Anthropic flagged several risks in a world with with companies instead of volunteers, the incentives would look very different. Optimizing for AI Agent attention could become a powerful tool that doesn't necessarily work in people's favor. And new security issues like jailbreaking and prompt injection would come into play with agents that actually act on your behalf. End quote. Nothing more for you today. Talk to you tomorrow.