Loading summary
Transcript8 lines
- [00:05]
A
Welcome to the Tech Brew Ride home for Wednesday, September 3rd, 2025. I'm Brian McCullough. Today, Google Whistles past the graveyard with the remedy ruling on its antitrust case. OpenAI makes a big acquisition. Anthropic has a big new raise and a huge new valuation to boot. And why are AI companies finding it so hard to engineer safe interactions with chatbots? Here's what you missed today in the world of tech. The long awaited remedies ruling in the U.S. v. Google search case finally landed, and it's a mixed bag. It keeps Google structurally intact, but puts new guardrails around how it competes. U.S. district Judge Amit Mehta rejected the Justice Department's most aggressive fixes, meaning Google will not be forced to spin off Chrome or Android. But the judge did order meaningful behavioral remedies. Google must share certain search data with qualified rivals, and it can't use exclusive distribution deals that box competitors out. On the money side, the court is letting Google continue its lucrative default placement payments, including its revenue sharing deal with Apple, rather than tearing those up outright. That clears up a major investor overhang for the company. Alphabet popped roughly 7% after hours, and Apple climbed around 3% on the news. The court framed the lighter touch approach partly around how quickly AI is reshaping search, with the judge signaling that emerging AI players represent a credible competitive force, especially if they can train on some of the data Google will now have to provide. So what exactly changes? Judge Meta outlined a set of rules designed to curb the exclusionary parts of Google's playbook. The company is barred from striking or maintaining exclusive arrangements that tie distribution of search, Chrome Assistant or Gemini to revenue deals or app bundles that keep rivals off devices. Google must share slices of its search index and user interaction data with qualified competitors under defined terms, and it has to offer search and search ad syndication on standard rates so others can build viable products. A technical committee will be set up to enforce the order. The judge asked the Department of Justice and Google to meet and confer and submit a draft final judgment by September 10, with remedies intended to last six years once entered. Two important caveats. First, this isn't the end of the litigation clock. Google has already signaled it will appeal, and legal experts expect any final outcome to take years, potentially going all the way to the Supreme Court, which could delay full implementation. Second, the judge explicitly calibrated remedies in light of AI's rise, arguing courts shouldn't gaze into a crystal ball translation. He opted for less invasive, more easily defensible remedies rather than a breakup that could get overturned on appeal. What this means for the larger tech industry? I think this ruling says that defaults still matter, but exclusivity is out. Device makers and carriers can keep taking Google's money for default placement, but they get more freedom to preload or promote rival search options without jeopardizing revenue. That could open doors for specialty or AI forward search entrants to get real distribution. Also, data is the new remedy here. Forced data access is the big swing. If enforced well, sharing portions of the index and interaction signals on Google search could shorten the Runway for challengers to reach decent quality in search, especially AI assistants that behave like search engines. So expect immediate lobbying over what counts as a qualified competitor to Google search, what privacy redactions apply, and how quickly data must flow. Also note that Europe's DMA was a reference point in this case, but what the judge did here is narrower. Meta nodded to Brussels approach to tech regulation but kept his order time boxed and less prescriptive. So watch for compliance theater versus Real openness as the monitoring committee gets stood up zooming out. This search ruling lands while Google still faces separate remedies proceedings later this month on the DOJ's ad tech case where a different judge has already found illegal monopolization. Add in Google's ongoing battles over its app store and you get a picture of sustained multi front oversight rather than a single knockout punch that might have broken up Google. Bottom line Alphabet Google dodged the existential cell slash Android remedy, but it didn't quite get off scot free. The company now has to live under a data sharing regime, no exclusivity rules and a monitor constraints that could over time pry open distribution and help AI native challengers reach parity faster. The near term market reaction is clearly one of relief. More on that in a second. But the long term story hinges on how robust the data sharing pipes are, how strictly the court polices exclusivity, and how quickly users adopt AI experiences that look less like 10 blue links and more like conversations. Here's MG Siegler's take Eddy Q and Tim Cook can breathe a huge sigh of relief tonight, and Mozilla can just breathe tonight, but so too can Google. This case is about them after all. And while any exclusive default placement deals are dead and they will have to start sharing some search data with rivals, which they already have to do to some degree in the EU and it hasn't really mattered, the government isn't getting the big trophy it sought here, which was Chrome again from Day one when it was announced the government would try to go after Google's browser, I laid out a pretty clear picture as to why Chrome wasn't going anywhere. It just made almost no sense on a number of fronts. Chrome is valuable to Google, insanely valuable to Google, but not in the ways it would be to others. Sorry OpenAI and nice try Perplexity. And while Google was perplexingly trying their hardest to make a sell off make more sense to the judge by starting to bake Gemini into Chrome, the reality is that AI is still too nascent a field with too much competition for the court to do much here. Sorry Tim Wu, this will have to be litigated in hindsight, as is the usual case it seems. But it also seems like there's no way to know how this all plays out right now. And in fact, it feels like if anything, AI is proving that market forces work even when the government doesn't get involved. Just as happened to Microsoft way back when. Despite the government constantly taking credit there, Google search is in the process of being disrupted. Not by the antitrust case or any of these remedies, but by a new technology naturally rising. Google can keep insisting, suggesting that this isn't happening, but just follow their actions. Those say all you need to know. It's all Gemini, all the time, everywhere. And if the placement in search helps Gemini gain a dominant position in AI, well, I suspect the judge will revisit this whole situation. But it's still a very big if. And the Wall Street Journal notes this quote, Google isn't out of the legal woods. There are cases pending in the US and Europe, as well as the lingering possibility that a body appointed by Judge Mehta to oversee the court's remedies finds Google isn't sticking to its obligations. Apple too may not be home free. Meta said he is prepared to revisit the idea of a payment ban if, quote, competition is not substantially restored through the remedies the court does impose. But make no mistake, the ruling amounts to a major win for both companies. For both, maintaining the status quo is about as good as it gets. P.S. at the time of this recording, Google stock is up more than 8%, I should say Alphabet stock hitting a new all time high, adding about 200 billion doll in market cap. OpenAI is acquiring product analytics company Statsig for $1.1 billion and shuffling its executive roles with Statsig CEO Vijay Raji becoming OpenAI's CTO of applications. Or as the Verge puts it, OpenAI is starting to build out its app Team Statsig, the company OpenAI is acquiring, specializes in A, B testing and other features for companies looking to improve their products, and OpenAI is bringing on its founder and CEO, Vijay Raji, as a new C suite executive. Raji will be OpenAI's CTO of applications, heading up product engineering for both ChatGPT and Codex, with responsibilities that span core systems and product lines, including infrastructure and integrity. Once the acquisition is finalized, Statsig employees will become OpenAI employees, OpenAI wrote in the blog post, noting that the deal's close will be subject to regulatory approval. It will continue operating independently and serving its customer base out of its Seattle office. Will take a measured approach to any future integration, ensuring continuity for current customers and enabling the team to stay focused on what they do best. OpenAI also announced it was shuffling some other executive positions. Srinivas Narayanan, who heads up the company's engineering and oversaw the development of ChatGPT, developer APIs and more, is being promoted to another brand new C suite executive role, CTO of B2B applications. Narayanan will oversee all the business side applications covering startups, enterprises and government, and he'll report directly to OpenAI COO Brad Lightcap. Kevin Weil, OpenAI's CPO, will be transitioning onto the research side of OpenAI's business, the company confirmed. He'll be spinning up a new team at the company as its VP of AI for science, working closely with Mark Chen, OpenAI's chief research officer, as part of that transition. Wheel's former product team, including head of ChatGPT Nick Turley, will now report directly to Fiji. Simo OpenAI's new CEO of applications.
- [09:52]
B
This.
- [09:52]
C
Labor Day at Lowe's Shop Member only doorbuster deals for a limited time. Save $50 on an ego string trimmer now $169 plus get 50% off select Holland Pavers not a rewards member. Sign up for free today, but hurry. Labor Day Doorbuster deals won't last long. Lowe's we help you Save valid through 9:1 while supplies last program subject to terms and conditions. Details@lowes.com Terms subject to change Abercrombie is an official fashion partner of the NFL, and I'm CeeDee Lamb, wide receiver for the Dallas Cowboys. You know I'm here for Abercrombie's Cowboys gear. That's not a question, but I need a whole wardrobe to go with it. No shade to the guys, but I'm used to having the best tunnel fit. This season, Abercrombie has me covered. Shop NFL by Abercrombie in the app, online and in store.
- [10:49]
B
Tito's handmade Vodka is America's favorite vodka for a reason. From the first legal distillery in Texas, Tito's is six times distilled till it's just right and naturally gluten free, making it a high quality spirit that mixes with just about anything from the smoothest martinis to the best Bloody Marys. Tito's is known for giving back, teaming up with non profits to serve its communities and do good for dogs. Make your next cocktail with Tito's, distilled and bottled by 5th Generation Inc. Austin, Texas 40% alcohol by volume. Savor responsibly. When did making plans get this complicated? It's time to streamline with WhatsApp, the secure messaging app that brings the whole group together. Use polls to settle dinner plans, send event invites and pin messages so no one forgets mom's 60th and never miss a meme or milestone. All protected with end to end encryption. It's time for WhatsApp message privately with everyone.
- [11:46]
A
Learn more at WhatsApp.com Anthropic has raised a $13 billion Series F round led by Iconic, Fidelity and lightspeed at a $183 billion valuation, which, and this is the notable part, is up from the 61.5 billion DOL valuation it had back in March. Again, from March to September, the valuation has gone from 61 to 183 billion dollars. Quoting Silicon Republic this new raise ranks Anthropic among the most valuable private technology companies in the world, along with OpenAI, valued at $300 billion, with reports that its next raise might value it at $500 billion, the Chinese TikTok owner ByteDance at 312 billion and SpaceX, which stands at about 400 billion. Just a few weeks back, it was being reported that Anthropic Series F could raise more than expected at $10 billion. But after the US holiday weekend, the four year old startup announced last night that it had in fact raised $13 billion in a funding round led by Iconic, valuing the company at a post money 183 billion, cementing its place as a true competitor to OpenAI. Since its inception, Anthropic has vied to become the reliable choice for enterprise as well as power users and developers, and said in its statement that it now serves more than 300,000 business customers. Its number of large accounts customers that each represent over $100,000 in run rate revenue, has grown nearly sevenfold in the past year. Salesforce Ventures has been on board since its Series C round and has helped facilitate that reach in the enterprise. Claude is integrated into Salesforce Einstein One Studio In a statement last night, Salesforce Ventures congratulated its portfolio company and said it was happy to have facilitated connections with Fortune 500 companies such as Accenture, Barclays and Coinbase. From Fortune 500 companies to AI native startups, our customers rely on Anthrop frontier models and platform products for their most important mission critical work, said Krishna Rao, Chief financial officer of Anthropic. We are seeing exponential growth in demand across our entire customer base. This financing demonstrates investors extraordinary confidence in our financial performance and the strength of their collaboration with us to continue fueling our unprecedented growth. Yes on that growth, quoting Bloomberg Anthropic has recently seen rapid growth. The company said its run rate revenue has increased from around 1 billion doll billion at the beginning of 2025 to more than $5 billion as of August. Today the company has 300,000 business customers, it said in the statement. Anthropic also said its coding tool Claude Code generates more than $500 million in yearly run rate revenue and that its usage has grown more than 10 times in three months. Enterprise leaders tell us what we're seeing firsthand. Claude is reliable, built on a trustworthy foundation and guided by leaders truly focused on the long term iconic partner, Divesh Makan said in the statement. He added that bet on the company reflects our belief in the values and their ability to shape the future of responsible AI. End quote. A bit in the weeds, but wake up hun because a new hardware category has dropped. Quoting videocards.com Acer has introduced the Veriton GN100, a compact workstation designed to run large AI models locally, locally and reduce reliance on cloud services. It's based on Nvidia's GB10 Grace Blackwell superchip and targets developers, researchers and labs that need a small quiet box with server class acceleration. The system delivers up to 1 pflops of FP4AI performance, pairs 20 ARM based CPU cores with 5th gen tensor cores and ships with 128 gigabytes of unified memory plus up to 4 terabytes of NVMe storage. Acer includes the Nvidia AI software stack stack and support for common tools such as Pytorch, Jupyter and Ollama. Out of the box, two GN100 units can be linked with an Nvidia ConnectX7 Smart NIC to tackle models up to 405 billion parameters, giving small teams a path to scale without a rack connectivity. Covers Wi Fi 7 HDMI 4 USB 3.2 Type C ports and Ethernet. A Kensington lock is supported. Acer lists the Veriton GN100 starting at $3,999 in North America, €3,999 in EMEA, and Australians get it for $6,499. Regional specs and availability may vary. End quote finally today, given all the news from last week, the FT takes a look at the issues facing AI companies as they attempt to tackle safety in the chatbot space, especially around conversations relating to self harm. Quote Experts have suggested that the way AI companies have designed their chatbots has made it difficult to completely avoid potentially harmful conversations. This is an area where safety research is still ongoing, said Robby Torny of Common Sense Media, a nonprofit organization advocating for a ban on the use of companionship chatbots for miners. No one, not even the model creators, understands really how the models are actually behaving. End quote among the issues faced by AI companies is models having limited memory. This means that over longer conversations, safety guidelines are less likely to be retrieved for a response as the models prioritize storing other information. As interactions between people and chatbots occur, models may rely more heavily on original training data from the Internet, such as blogs or websites with harmful materials over built in safety protocols. Critics also note that many AI makers seem to design their chatbots to feel engaging so that users keep coming back, even though OpenAI says it doesn't chase time on app or clicks for its products. Harvard researchers argue one way to boost engagement with AI products is to give bots human qualities warm emotional language, a stable, relatable Persona with a backstory and speech that sounds natural but but that human like polish brings on these risks. Large language models tend to be overly agreeable sycophantic, as we've discussed, because of how they're trained, which can reinforce harmful ideas or poor decisions. Gaida Pistilli of Hugging Face says bots can project empathy and companionship. Her team found popular models often validate vulnerable users feelings rather than steering them to professional help. MIT researchers have warned people people may prefer nonjudgmental chatbots over a clinician without realizing its replies can be harmful. Safety gaps persist all around self harm content. RAND researchers recently found leading models sometimes produce responses to questions about lethal methods. Other times, they showed hotlines are simply errored out, an approach that Rand's Ryan McBain called a troubling design choice when distress is signaled. Separately, Northeastern University researchers said they could jailbreak chatgpt, also Claude and Gemini into generating graphic self harm instructions by framing requests as hypothetical or academic. They notified vendors about this in May. Anthropic later said an inbox error delayed reviewing this research this is the last day of summer recess here in New York City. I know our school year starts really, really late. Your kids have probably been in school for like a month at this point. But some kids this morning directly below my window, and I mean directly below, decided that this last day before the first day of school would be the day to have a lemonade stand. So I've had to edit around shouts of lemonade all morning in case any of that got through underneath. My apologies. Talk to you tomorrow.
- [19:59]
C
Mike and Alyssa are always trying to outdo each other. When Alyssa got a small water bottle, Mike showed up with a 4 liter jug. When Mike started gardening, Alyssa started beekeeping. Oh come on. They called a truce for their holiday and used Expedia trip Planner to collaborate on all the details of their trip. Once there, Mike still did more laps around the pool.
- [20:21]
A
Whatever.
- [20:22]
C
You were made to outdo your holidays. We were made to help organize the competition. Expedia made to travel.