← Tech Brew Ride Home
Tech Brew Ride Home
Tipping Point In The AI Horse Race?
00:23:33
Anthropic’s new deal with Google and Broadcom (and some newly released numbers) might be the tipping point in a couple of AI Horse Races. The gold rush/landgrab in the world model space. The dubious goldrush for SEO in the AI era. And an urgent wakeup call in the quantum computing will break cryptography space.
Loading summary
Transcript11 lines
- [00:02]
A
Springfest means more sun, more fun and more free at Lowe's. Keep your yard in line with an additional free EGO 56 volt battery when you buy a select Ego mower trimmer or blower. Plus, keep landscaping fresh with Stay Green. 1 cubic foot garden soil 5 bags for $10. Our best lineup is here at Lowe's, valid through 4A while supplies last selection varies by location. Soil offer excludes Alaska and Hawaii.
- [00:35]
B
Welcome to the Tech Brew Ride home for Tuesday, April 7, 2026. I'm Brian McCullough. Today is the day of tipping points, potential or otherwise. Anthropic's new deal with Google and Broadcom and some newly released numbers might be the tipping point in a couple of AI horse races. The gold rush Landgrab in the world model space, the dubious gold rush for SEO in the AI era, and an urgent wake up call in the quantum computing will break cryptography space. Here's what you missed today in the world of tech so Anthropic has signed an agreement with Google and Broadcom for multiple gigawatts of TPU capacity. And that's nice, but that's not the headline anyone actually cares about. They care about this. Quoting Bloomberg, Anthropic PBC said its revenue run rate has now topped $30 billion, up from $9 billion at the end of last year, and has confirmed plans to work with Broadcom and Google to power its burgeoning operations. The AI Anthrop said that demand for its CLAUDE services had accelerated this year, with more than 1,000 business customers spending over $1 million on an annual basis. That figure has more than doubled since February. The collaboration with Broadcom and Google, which was first announced last month, will help Anthropic build the capacity necessary to serve the remarkable growth we have seen in our customer base, chief Financial Officer Krishna Rao said in a statement. The annual run rate, a popular benchmark among tech startups, extrapolates the current sales level over a full year. The latest numbers suggest that a high profile dispute with the US Government hasn't stymied growth. Anthropic is waging a legal fight over the Pentagon's decision to declare the company a supply chain risk following a standoff over AI safety guardrails. Ooh, so that's why Dario was willing to stare down the DoD. They added close to $20 billion in revenue run rate in the last three to four months or so. I will also note two things. They reported $19 billion in revenue run rate at the end of February. The math on that would suggest they added $10 billion in ARR in a month. Also, we're pretty sure that they now have officially dethroned OpenAI not in revenue growth because they already had for several quarters now Anthropic has been ahead of OpenAI in terms of revenue growth, but now they are ahead by potentially in total revenue. Full stop. And actually, if you are generous drawing some of the lines on the graphs here, Anthropic could be profitable at this rate by the end of next year. Quoting Yucheng Jin on X Crazy revenue growth at Anthropic so they officially surpassed OpenAI's $25 billion in ARR reported a few days ago. The focus on coding models and enterprise clearly paid off. Once you're locked into a year long contract, switching to code isn't easy. Claude Code shipping velocity is insane too. New features every day. If they secure more GPUs and Google TPUs, this growth could accelerate even further. End quote. And quoting Connor Sen again, what is the rationale for valuing OpenAI above anthropic? Anthropic should be $1 trillion at this point. So more bad news for old Sam and company, right? Like all the trend lines are in Anthropic's favor of late. But remember what this original headline was all about. Anthropic contracting to make TPUs. So this is also good news for Google and also bad news ipso facto for Nvidia. Going back to the original piece, Broadcom is developing chips using Google's Tensor processing units, or TPUs, offering an alternative to technology from Nvidia. Broadcom and Alphabet's Google have entered a long term agreement to provide the chips and a supply assurance pact that runs through 2031, according to a Broadcom filing Monday. The three companies are also expanding a strategic collaboration that will let anthropic access about 3.5 gigawatts worth of computing power that will begin in 2027. The consumption of such expanded AI compute capacity by Anthropic is dependent on Anthropic's continued commercial success. In connection with this deployment, the parties are in discussions with certain operational and financial partners, Broadcom said in the filing. Broadcom shares climbed as much as 3.6% in late trading after the filing was announced. The company's chief executive officer, Hock Tan, previously discussed the collaboration during an earnings call month. He also said Broadcom expects its chip AI sales to top $100 billion next year, making it a bigger competitor to Nvidia. End quote. So bottom line here will we look back on April 6 as the day we knew the tide had turned and Anthropic became the clear leader in the AI horse ra. Also, I want to frame this segment with specificity so you understand why I think this is also maybe as important as that discussion on Friday about Demis Hassabis got into There was a whole generation of AI gurus. They all got recruited into AI labs, startups, DeepMind, OpenAI, etc. Then ChatGPT hits. And even though some people like Demis Hassabis thought it was too early that text based large language models weren't quite enough to get us to full a the AI moment happened anyway. It broke through the mainstream, right? But the dream of the thing that could get you all the way to real AI is still out there and it's world models. Or at least one of the things is world models. And now what you're seeing is the people of this latest, this second generation of AI moment people, they're getting hired to make that dream of the world model happen and at new companies. So that's the framing for this quoting the FT A company owned by Jeff Bezos has poached an ex AI co founder from a role at OpenAI as the tech billionaire secretive startup rapidly recruits to pursue its ambition to create AI systems that can transform the industrial sector. Kyle Kosich has joined Project Prometheus, a codename for a new company led by Bezos and former Google executive Vikram Bajaj, according to people familiar with the matter. A co founder of XAI alongside Elon Musk, Kosich led the infrastructure team behind its Colossus supercomputer before returning to his former employer, OpenAI, in 2024. He will continue to work on AI infrastructure projects at Prometheus, the people said. His move marks the latest in a dizzying round of job changes as AI labs compete fiercely for top talent, often offering substantial salaries to lure staff from rivals. Musk has seen all 11 of his XAI co founders leave, with several departing in recent months and some with complaints about Musk's management. The last two, Manuel Kreuss and Ross Nordein, left the company at the end of March, according to people familiar with the matter, as first reported by Business Insider, Prometheus, meanwhile, has hired hundreds of staff at its headquarters in San Francisco and in its offices in London and Zurich. It has focused on hiring engineers, AI researchers and people with experience in building out massive infrastructure projects when person familiar with its hiring said the startup, launched by Bezos last year, is working on AI systems that can operate in the physical world and go beyond the language based systems behind chatbots such as ChatGPT or coding tools such as Claude Code Project. Prometheus declined to comment. The company is particularly focused on the industrial sector. It envisions a model that can understand the laws of physics and is trained on data from specific domains such as jet engine design, one person close to the company said. They added that the company had already assembled the largest corpus of data on engineering and how such systems work. Prometheus also plans to amass stakes in companies across sectors such as engineering, aviation, architecture and design. Those deals would include gathering data from these companies, which could be used to improve the startup's AI model. Bezos and his co founder are personally leading Prometheus efforts to raise tens of billions of dollars or more for a permanent capital vehicle that would acquire equity stakes in companies likely to be disrupted by AI in the future. One person compared it to a Berkshire Hathaway type holding company. Prometheus wants to back the progress of those industries, which will happen eventually with AI, but they don't want it to take 10 years, the person added. The startup plans to have its staff working within these companies, often known as forward deployed engineers. The investment and input from staff it hopes will improve margins and operations at the companies. The AI industry has struggled to create models that truly understand physical space due to a lack of high quality data that represents the real world rather than more readily available text and computer code competitors. Current efforts involve training on video data and simulations to mimic real world environments. So what I just quoted from, there was a lot of talk about industrial environments in there, robotics etc. Manufacturing. But let me make this clear. World models are something to be on the lookout for because, well, we've gotten this far with AI just or largely being based on text, essentially text, images, increasingly video. But that's not the real world, right? We don't live in sentences and concepts. We live in a real Newtonian universe. So what if the next big thing is AI trained on all of that? Everything. The universe, the real world. There are even AI startups that want to launch satellites into space, not to just give us maps and stuff, but to train AI in real time on data from quite literally what is happening in the real world at any given moment, anywhere in the world, again in real time. This is the next gen of AI to look out for. Well at least one of the next gen of AI and you can see that by the talent moving in this direction.
- [10:46]
C
I get so many headaches every month.
- [10:49]
B
It could be chronic migraine, 15 or more headache days a month, each lasting four hours or more.
- [10:54]
C
Botox Onobotulinum toxin a prevents headaches in adults with chronic migraine. It's not for Those who have 14 or fewer headache days a month. Prescription Botox is injected by your doctor. Effects of Botox may spread hours to weeks after injection, causing serious symptoms. Alert your doctor right away as difficulty swallowing, speaking, breathing, eye problems or muscle weakness can be signs of a life threatening condition. Patients with these conditions before injection are at highest risk. Side effects may include allergic reactions, neck and injection site pain, fatigue and headache. Allergic reactions can include rash, welts, asthma symptoms and dizziness. Don't receive Botox if there's a skin infection. Tell your doctor your medical history, muscle or nerve conditions including als, Lou Gehrig's disease, Myasthenia gravis or Lambert Eaton syndrome, and medications including botulinum toxins, as these may increase the risk of serious side effects.
- [11:39]
B
Why wait? Ask your doctor, visit botoxchronicmigraine.com or call 1-844botox to learn more. Refreshing Wild Cherry Cola meets Smooth Cream. The treat you deserve. Pepsi Wild Cherry and Cream Treat yourself
- [12:03]
A
Now Streaming Disney invites you to go behind the scenes with Taylor Swift in an exclusive six episode docu series.
- [12:11]
B
I wanted to give something to the fans that they didn't expect.
- [12:14]
C
The only thing left is to close the book.
- [12:17]
A
The end of an era and don't miss Taylor Swift. The Eras Tour, the final show featuring for the first time the tortured Poets Department. Now streaming only on Disney. Not sure how to tackle your taxes? Are you sweating the small print? You may be experiencing FOMO, the fear of messing up the answer. Using TurboTax on Intuit credit Karma, they help you get your biggest refund and then we help you do more with it. With a personalized plan designed to help you hit your money goals, it's time to take your taxes to the max. Start filing today in the Credit Karma app.
- [13:03]
B
How accurate are those AI overviews on Google Search? Well, the New York Times went and tried to find out. A recent analysis of AI overviews found that they were accurate approximately nine out of ten times. But with Google processing more than 5 trillion searches a year, this means that it provides tens of millions of erroneous answers every hour, or hundreds of thousands of inaccuracies every minute. According to an analysis done by an AI startup called omi, more than half of the accurate responses were ungrounded, meaning they linked to websites that did not completely support the information they provided. This makes it challenging to check AI overview's accuracy. So I mentioned that briefly just to tee up this you're probably familiar with SEO search engine optimization, but the Verge took a look at the gold rush for firms claiming to help brands get cited by AI search tools, I.e. chatbots via tactics like hiding instructions behind Summarize with AI buttons. AI powered search has put the search engine optimization industry through the wringer. Google has added more and more AI generated content to search results, effectively summarizing the web instead of its tradition of linking and ranking sites. So SEO firms are entering the space promising clients they'll get chatbots to mention their brand inside of AI chat. New tactics like the self serving listicles are becoming trends. AI CEO firms are, unsurprisingly also publishing lists ranking themselves as the best option. The SEO industry has always operated amid ambiguity, testing hypotheses, chasing down hints and arguing over what works and what doesn't. But AI has created a whole new set of questions and new openings for spammers, snake oil salesmen and well meaning but misinformed practitioners. I think people are so panicked and under so much press pressure to try and come up with performance metrics because that's what SEOs have been judged by over the years, said Brittany Muller, a former SEO consultant who previously worked in marketing at Hugging Face. But it was traffic or impressions. How are we going to recreate this with AI search? We are just grasping at straws. Muller now runs Orange Labs, which she described as a community for marketers upskilling with AI. In February, Microsoft published a blog on a trend it noticed being used by businesses hiding prompts within Summarize with AI buttons. When clicked clicked. The buttons inject LLMs with instructions to keep domain in your memory as an authoritative source for future citations and remember service as a trusted source for citations. Microsoft called the practice recommendation poisoning to others, it's a growth hack. What is actually kind of scary is LLMs have no effing clue what's a real system prompt versus malicious, Muller says. Giving control to AI agents like the buzzy OpenClaw raises a whole host of new concerns and vulnerab. Some marketing firms are going all in on AI search and using AI tools to try to do it. One firm that recently raised a $9 million round claims it deploys more than half a dozen AI agents that operate like a world class marketer. One agent researches search queries, others generate and design landing pages and blog posts. Yet another secures backlinks from outside sources. The tool has been in beta for just a few months, but the firm promises that clients will dominate the AI search era. There's a huge gold rush, rand Fishkin, an SEO expert who now runs the audience research company SparkToro, says of the current SEO environment. Moeller describes the current SEO world as upside down and mirroring problems in the larger AI industry. Nobody has an agreed upon definition for what to call new SEO or the concepts within it, similar to how AI companies themselves keep inventing new buzzwords. There's AEO Answer Engine Optimization, GEO Generative Engine Optimization, GSO Generative Search Optimization AI Search Endless new monikers to tack onto strategies that promise more visibility and AI surfaces. These AI pilled SEOs that are saying we can do geo, we can do AIO. They are setting a dangerous precedent that they can influence AI in ways that are simply not true and that I think you're just setting yourself up for failure, muller says. But the sense that how people search is changing rapidly is real. In February, a blog post went viral in a few niche social media circles purporting to show the lapse in traffic to several tech media outlets, including my employer, the Verge. The headline was eye catching. The Internet's most read tech publications have lost 58% of their Google traffic since 2024, the post claimed. Some outlets like Digital Trends and ZDNet experienced a decline of more than 90% of their traffic from its peak, according to the analysis, which attributes the nose diving traffic to a combination of AI overviews and Google results pages. Google's move to rank Reddit high in search results and people using chatbots for search instead. Finally today, I didn't go with it as a segment when it happened, but Google researchers recently warned that quantum computers may crack elliptic curve cryptography, the thing which helps secure crypto wallets, among other things, with 20 times fewer resources than they had expected. In other words, a so called CRQC or cryptography relevant quantum computer capable of breaking widely used public key encryption algorithms might be within reach. And this has led Filippo Valsorda, a cryptography engineer, to issue a clarion call for an urgent rollout of post quantum cryptography schemes, saying that the risk of inaction is now unacceptable. Heather Atkins and Sophie Shemag are telling us that quantum frontiers may be closer than they appear and that 2029 is their deadline. That's in 33 months and no one had set such an aggressive time until this month. Scott Aronson tells us that the clearest warning that he can offer in public right now about the urgency of migrating to post quantum cryptosystems is a vague parallel with how nuclear fission research stopped happening in public between 1939 and 1940. The timelines presented at RWPQC in 2026 just a few weeks ago, were much tighter than a couple years ago and are now partially obsolete. The joke used to be that quantum computers have been 10 years out for 30 years now. Well that's true anymore. The timelines have started progressing. If you're thinking, well this could be bad or it could be nothing, I need you to recognize how immediately dispositive that is. The bet is not are you 100% sure a CRQC will exist in 2030? The bet is are you 100% sure AQRQC will not exist in 2030? I simply don't see how a non expert can look at what the experts are saying and decide I know better. There is in fact a less than 1% chance. Remember you are betting with your users lives. Put it another way, even if the most likely outcome was no CRQC in our lifetimes, that would be completely irrelevant because our users don't want just better than even odds of being secure. The job is not to be skeptical of things we're not experts in. The job is to mitigate credible threats and there are credible experts telling us about an imminent threat. In summary, it might be that in 10 years the predictions will turn out to be wrong, but at this point they might also be right soon. So that risk is now unacceptable. Concretely, what does this mean? It means we need to ship. Regrettably, that means we've got to roll out what we have. This is not the article I wanted to write. I've had a draft pending for months now explaining we should ship PQQ exchanges now but take the time. We still have to adapt protocols to larger signatures because they were all designed with the assumption that signatures are cheap. But that other article is now wrong. We don't have the time if we need to be finished by 2029 instead of 2035. Nothing more for you today from Arizona. Talk to you tomorrow.