Build by AI

Anthropic's Near-Trillion Dollar Moment I 29th May

Share episode

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 30:28
When Anthropic closed a $65 billion funding round at a $965 billion valuation, it wasn't just another startup milestone - it was a glimpse into how AI companies are rewriting the rules of Silicon Valley economics. We dive into Claude Opus 4.8's benchmark-beating performance, explore Apple's rumored Siri overhaul for iOS 27, and connect the dots on what happens when AI budgets start shrinking across enterprise. The future of AI is getting expensive, competitive, and surprisingly honest about its mistakes.
SPEAKER_01

I was making coffee this morning when the notification popped up on my phone. Anthropic $65 billion $965 billion valuation. I literally had to put the coffee down and read it again because those numbers just don't compute in normal human terms.

SPEAKER_00

I had the exact same reaction. I was scrolling through the news and saw that headline and thought there had to be a typo. Like we're talking about a company that's basically worth more than most countries' GDP at this point.

SPEAKER_01

And then you realize, you know, this isn't even their IPO. This is just their last private round before they go public. We're watching the birth of what might be the most valuable tech company in history.

SPEAKER_00

Right. And it's happening at the exact same time they're releasing Claudopus 4.8, which is apparently beating GPT 5.5 on most benchmarks. The timing feels very deliberate.

SPEAKER_01

Oh, it's absolutely deliberate. When you're about to ask public investors for a trillion dollar valuation, you better have the receipts to back it up.

SPEAKER_00

And the crazy thing is, this is happening while everyone else is talking about cutting AI budgets and being more cost conscious. There's something fascinating about that disconnect. You're listening to Build by AI, the daily show where we break down the AI news that actually matters. I'm Alex Shannon.

SPEAKER_01

And I'm Sam Hinton. Today we're diving deep into Anthropic's historic funding round, their new model that's more honest about its mistakes, and some interesting rumors about Apple's plan to turn Siri into a ChatGPT competitor.

SPEAKER_00

Plus we'll look at how enterprise AI budgets are getting tighter and what that means for startups trying to compete with the big players.

SPEAKER_01

It's going to be a packed show, so let's jump right in.

SPEAKER_00

For context, that puts them just $35 billion away from becoming the first AI company to hit a trillion dollar valuation. And according to multiple reports, this is expected to be their final private raise before pursuing an IPO.

SPEAKER_01

Yeah, this is absolutely massive. To put those numbers in perspective, that $965 billion valuation makes Anthropic worth more than Tesla, more than Meta, more than most of the companies in the S P 500. We're talking about a company that was founded in 2021 and is now worth nearly a trillion dollars.

SPEAKER_00

That timeline is just wild to think about. So what's driving this kind of valuation? Is it purely hype, or is there something fundamentally different about how investors are looking at AI companies now?

SPEAKER_01

I think it's a combination of factors. First, the revenue growth in enterprise AI is unlike anything we've seen before. Companies are spending billions on AI infrastructure and tools, and that market is only going to get bigger. But more importantly, I think investors are betting that whoever wins the foundation model race is going to have a monopoly-like advantage for years to come.

SPEAKER_00

But here's what I'm wondering. Can any company actually justify a trillion dollar valuation in the AI space right now? I mean, even OpenAI, which has been the market leader, isn't at these numbers yet.

SPEAKER_01

That's a great point. And honestly, I'm a bit skeptical. The AI market is still so young and volatile. We've seen models leapfrog each other in performance every few months. Just because Claude is winning benchmarks today doesn't mean they'll be winning them six months from now. A trillion dollar valuation assumes they're going to maintain that competitive edge indefinitely.

SPEAKER_00

Right. And there's also the question of what happens when the AI bubble eventually corrects. We've seen this pattern before with other tech trends, massive valuations followed by reality checks when the market matures.

SPEAKER_01

Exactly. But here's the thing. If Anthropic can successfully IPO at or near this valuation, it sets a new benchmark for how the market values I companies. Every other startup is gonna point to this deal when they're raising their next round.

SPEAKER_00

Which could create this cascading effect where AI valuations across the board get inflated. But let me ask you this. What if they're right? What if foundation models really do become the new operating systems of the digital economy?

SPEAKER_01

If that's true, then a trillion dollar valuation might actually be conservative. Think about Microsoft's market cap. They're worth over three trillion dollars now, largely because Windows became the foundation that everything else was built on. You know, if Claude becomes that foundational layer for AI applications, then yeah, the numbers start to make sense.

SPEAKER_00

But that's a huge if. And it assumes that AI models will have the same kind of network effects and switching costs as traditional platforms. I'm not convinced that's the case. It seems like customers can switch between AI models pretty easily.

SPEAKER_01

That's true today. But what about when companies start building their entire workflows around specific AI capabilities? Once you've trained your team on Claude's interface, integrated it into all your systems, built custom workflows around its strengths, switching becomes much more expensive.

SPEAKER_00

Okay, that's fair. So what should people be watching for as this moves toward an IPO? What are the key indicators that will tell us whether this valuation is sustainable?

SPEAKER_01

Revenue growth is going to be crucial. They'll need to show not just that they're growing, but that they're growing sustainably, and that their customers are sticky, and they'll need to demonstrate some kind of competitive moat, whether that's through their safety research, their model performance, or their enterprise relationships.

SPEAKER_00

Customer retention will be huge. If they can show that companies who adopt Claude stick with it and expand their usage over time, that's the kind of metric that justifies premium valuations.

SPEAKER_01

Absolutely. And I think the enterprise focus is smart. Consumer AI is still pretty volatile and price sensitive. But enterprise customers who see real productivity gains from AI tools, they'll pay premium prices and sign long-term contracts.

SPEAKER_00

The other thing to watch is how the competitive landscape evolves. Google isn't going to just sit back and let Anthropic dominate enterprise AI. Neither is Microsoft or Amazon. The question is whether Anthropic can maintain their lead long enough to build a sustainable business. Speaking of Anthropic's competitive edge, let's talk about their new model release. They're shipping Claude Opus 4.8, which they're calling a modest but tangible improvement over previous versions. But the interesting thing here is that it's not just about raw performance. They're emphasizing that this model is more honest and better at acknowledging its mistakes.

SPEAKER_01

And that honesty angle is actually a huge deal. One of the biggest problems with current AI models is overconfidence. They'll give you completely wrong answers, but deliver them with total certainty. If Claude can actually say, I'm not sure about this, or I might be wrong, that's a game changer for enterprise adoption.

SPEAKER_00

The technical details are pretty impressive too. According to the reports, Claude Opus 4.8 is beating GPT 5.5 and Gemini 3.1 Pro on most benchmarks, and it's catching coding errors four times more often than its predecessor. That's the kind of improvement that actually matters for real-world applications.

SPEAKER_01

Yeah. That coding error detection is huge. If you're a developer using AI to help write code, you want a model that's going to flag potential problems, not just generate code that looks correct, but has subtle bugs. That 4X improvement could be the difference between AI being a helpful tool versus a liability.

SPEAKER_00

But here's what I'm curious about. How do you actually measure honesty in an AI model? That seems like a pretty subjective quality to benchmark.

SPEAKER_01

And honestly, uh, I wish we had more details on their methodology. My guess is they're looking at things like calibration. Does the model's confidence level actually correlate with accuracy? And probably testing edge cases where the model should admit uncertainty rather than hallucinating an answer.

SPEAKER_00

Right. Because the real test isn't whether the model can be honest when it obviously doesn't know something. It's whether it can recognize the boundaries of its own knowledge in subtle situations where it might be tempted to guess.

SPEAKER_01

Exactly. And that's where I think Anthropic's background in AI safety research gives them an advantage. They've been thinking about these problems longer than a lot of their competitors. While OpenAI and Google were focused on pushing performance numbers, Anthropic was studying alignment and reliability.

SPEAKER_00

There's also this new dynamic workflows tool that lets you coordinate multiple AI agents for complex tasks. That sounds like Anthropic is moving beyond just building better models to building better AI systems.

SPEAKER_01

Exactly. And that's where I think the real value is going to be long-term. It's not just about having the smartest individual AI, it's about orchestrating multiple specialized agents that can work together. Think of it like having a team of experts rather than one generalist.

SPEAKER_00

The dynamic workflows approach is interesting because it addresses one of the fundamental limitations of current AI models. They try to be good at everything, which means they're not optimized for anything specific. But if you can break complex tasks into smaller specialized pieces.

SPEAKER_01

Right. You can have one agent that's really good at research, another that's optimized for data analysis, another for writing, and a coordinator that manages the whole workflow. Each piece can be more reliable because it's focused on what it does best.

SPEAKER_00

This actually makes me think about how human teams work. You don't expect one person to be the best researcher, analyst, writer, and project manager all at the same time. Division of labor works for humans, so why wouldn't it work for AI?

SPEAKER_01

And it probably makes the system more reliable overall. If one agent makes a mistake, the other agents can catch it. You get built-in error checking and quality control.

SPEAKER_00

So if you're a business evaluating AI tools right now, does this honesty factor change the calculus? Is it worth waiting for these more reliable models, or should companies be moving forward with current technology?

SPEAKER_01

I think it depends on your use case. If you're doing anything mission critical where errors could be expensive or dangerous, then waiting for more honest, reliable models makes sense. But for a lot of applications, content generation, initial research, brainstorming, current models are already good enough, and you're better off starting now and upgrading later.

SPEAKER_00

That's probably the right approach. And if Anthropic can maintain this focus on reliability and honesty while also staying competitive on pure performance, that could be their differentiator in an increasingly crowded market.

SPEAKER_01

It also plays into enterprise sales cycles. CTOs and IT directors aren't just looking for the most impressive demos, they want systems they can trust in production. Reliability sells better than flashiness in the enterprise world.

SPEAKER_00

Which brings us back to that valuation question. If Anthropic can position themselves as the reliable choice for enterprise AI, that's a valuable market position worth paying for. Alright, let's shift gears and talk about Apple. Now we've got early reports, and I want to emphasize these are early reports from a single source, suggesting that Apple is planning a major AI overhaul for iOS twenty seven twenty-seven, including a redesigned Siri experience and a standalone Siri app that's positioned to compete directly with ChatGPT and other AI assistants.

SPEAKER_01

Okay. If this is true, it's about time. Siri has been embarrassingly behind the curve compared to what we're seeing from OpenAI, Anthropic, and even Google. I mean, we're in 2026 and Siri still struggles with basic follow-up questions that ChatGPT was handling two years ago.

SPEAKER_00

The standalone app approach is interesting, though. Rather than just improving Siri within the existing iOS framework, they're apparently building something that can compete head-to-head with dedicated AI apps. What do you think about that strategy?

SPEAKER_01

I actually love it. Look, one of Apple's biggest advantages is that they control the entire ecosystem. If they can build an AI assistant that's deeply integrated with your calendar, your messages, your photos, your entire digital life, that's a level of personalization and utility that third-party apps just can't match.

SPEAKER_00

But there's also the privacy angle here. Apple has built their brand around privacy, and AI assistants typically require sending a lot of personal data to the cloud for processing. How do they square that circle?

SPEAKER_01

That's the million-dollar question. My guess is they're betting big on-device processing with their custom silicon. The M series chips are incredibly powerful. And if they can run a competitive AI model locally on your phone or laptop, that solves the privacy problem entirely.

SPEAKER_00

Right. And we've seen hints of this with their recent hardware releases. The neural engines in the latest chips are getting more sophisticated each generation. But can they really match the performance of cloud-based models like GPT or Claude running on massive server farms?

SPEAKER_01

Uh probably not in terms of raw capability, at least not initially. But they don't need to. If they can get to 80% of the performance while offering better privacy, um deeper integration in that classic Apple user experience polish, that might be enough to win over consumers.

SPEAKER_00

The timing is interesting too. iOS 27 would presumably launch in late 2026, which gives them time to see how the current AI landscape plays out and learn from everyone else's mistakes.

SPEAKER_01

Exactly. Apple's never been first to market with new technology, and they're usually second or third, but they nail the execution. Look at smartphones, tablets, smartwatches. If they can apply that same approach to AI assistants, ChatGPT and Claude might have some real competition.

SPEAKER_00

But here's what I'm wondering. Is Apple too late to this party? By the time iOS 27 launches, ChatGPT will have been in the market for like four years. Users will have established habits and workflows. How do you compete with that kind of head start?

SPEAKER_01

I think Apple's betting that most people still aren't using AI assistants regularly. Yes, the early adopters and tech enthusiasts are all over ChatGPT, but the average iPhone user probably hasn't integrated AI into their daily workflow yet. Apple could be positioning themselves for the mainstream adoption wave.

SPEAKER_00

That's a good point. And if they can make it feel seamless and natural, like AI features just work without you having to think about them. That could be more appealing to mainstream users than having to download separate apps and learn new interfaces.

SPEAKER_01

Right. Imagine if Siri could actually understand context from your entire device. Like you could say, remind me to follow up on that thing from earlier, and it would know you're talking about an email you read this morning. That kind of integration is impossible for third-party apps to replicate.

SPEAKER_00

The competitive dynamics here are fascinating too. If Apple can build a good enough AI assistant that's free and deeply integrated, that could seriously undercut the subscription model that companies like OpenAI and Anthropic are relying on. Of course, this is all still rumor and speculation at this point, but if confirmed, it signals that Apple is finally taking AI seriously as a competitive threat rather than just a feature enhancement.

SPEAKER_01

Siri's current state is becoming a liability when your voice assistant is noticeably dumber than what people can get for free from OpenAI. That's a problem for a company that prides itself on premium user experiences.

SPEAKER_00

The question is whether they can execute on this vision. Building competitive AI models is really hard, and Apple doesn't have the same depth of AI talent as Google or OpenAI. But they do have something those companies don't have total control over the user experience. Let's talk about something that might seem counterintuitive, given all these massive valuations, AI budget cutting. Early reports suggest that Glean, an enterprise AI search startup, has crossed $300 million in annual revenue by tripling year over year growth. And their major selling point is helping companies cut AI costs rather than just adding new AI capabilities.

SPEAKER_01

This is actually a fascinating trend that I think we're going to see more of. Right now, a lot of companies are spending money on AI tools without really understanding the ROI. You know, they're buying subscriptions to ChatGPT, Claude, multiple coding assistants, various automation tools, and they're discovering they have massive overlap and waste.

SPEAKER_00

So Glean is basically positioning themselves as the cost-conscious alternative to tech giants who are trying to sell companies on more and more AI tools.

SPEAKER_01

Exactly. And it's smart positioning because CFOs are starting to ask hard questions about AI spending. The honeymoon period where companies would buy any AI tool that promised productivity gains is ending. Now they want to see actual metrics and cost justification.

SPEAKER_00

That $300 million revenue number is pretty significant for what's essentially a search company. What does that tell us about the broader enterprise AI market?

SPEAKER_01

I think it tells us that though enterprise search is one of those unsexy but incredibly valuable AI use cases. Every large company has the same problem. Their employees can't find the information they need because it's scattered across dozens of different systems. If Glean can solve that with AI, companies will pay serious money for it.

SPEAKER_00

That's the kind of growth trajectory that venture capitalists love to see.

SPEAKER_01

Right. Because search is one of those foundational capabilities that every department needs. Once you prove value with one team, it's easy to roll out across the entire organization. That creates natural expansion revenue.

SPEAKER_00

But they're also competing with tech giants who have entered this category. How does a startup compete when Google, Microsoft, and others are building AI search into their existing enterprise products?

SPEAKER_01

Do the tech giants want to sell you their entire ecosystem. Microsoft wants you on Teams, SharePoint, Office 365, and Azure AI. Google wants you on Workspace and Google Cloud, but a lot of companies don't want to be locked into one vendor's ecosystem.

SPEAKER_00

So Glean is betting that companies will pay a premium for vendor neutrality and cost optimization over feature richness.

SPEAKER_01

Right. And in an economic environment where every company is being more careful about spending, that's probably a winning bet. It's easier to justify an AI tool that saves money than one that promises uncertain productivity gains.

SPEAKER_00

This also speaks to a broader maturation of the AI market. In the early days, everyone was focused on cutting-edge capabilities and impressive demos. Now enterprise buyers are asking more practical questions about integration, cost, and measurable business value.

SPEAKER_01

Exactly. And Glean's success suggests that there's a huge market for AI tools that solve specific, well-defined problems rather than trying to be everything to everyone. Enterprise search is boring, but it's a real problem that every company has.

SPEAKER_00

It also raises questions about the sustainability of the current AI startup funding environment. If enterprise customers are becoming more cost conscious, that could make it harder for new AI startups to achieve the kind of rapid growth that justifies massive valuations. Less focus on bleeding edge capabilities, and more focus on practical business value and cost management.

SPEAKER_01

Which ironically makes Anthropic's focus on reliability and the honesty even more strategic. If the market is moving towards Toward practical value over flashy demos, being the trustworthy AI provider could be a major competitive advantage.

SPEAKER_00

Alright, let's hit some rapid fire updates. We already covered Claude Opus 4.8's performance. But let's dig a bit deeper into that dynamic workflows feature that Anthropic released alongside it.

SPEAKER_01

Yeah, this is basically allowing you to coordinate swarms of AI subagents for complex tasks. Think of it like having a project manager AI that can delegate different parts of a complex problem to specialized agent workers.

SPEAKER_00

That sounds incredibly powerful for enterprise use cases. Instead of trying to prompt one AI to handle everything, you can break complex workflows into smaller, specialized tasks.

SPEAKER_01

Exactly. And it probably leads to better results because each subagent can be optimized for specific types of work rather than trying to be a generalist.

SPEAKER_00

The interesting thing is how this changes the user experience. Instead of having one conversation with one AI, you're essentially managing a team of AI workers. That's a pretty different mental model.

SPEAKER_01

Right, and it probably requires some learning curve for users. But if it delivers significantly better results for complex tasks, that learning curve might be worth it.

SPEAKER_00

This also feels like Anthropic is positioning themselves for a future where AI coordination and orchestration becomes as important as individual model performance.

SPEAKER_01

Which is smart because as models become more commoditized, the value shifts to how well you can integrate and coordinate them. It's not just about having smart AI, it's about having AI that works well together.

SPEAKER_00

Google's I.O. 2026 happened. An early report suggests they highlighted 12 major announcements, including updates to Gemini Omni and something called Gemini 3.5 Flash.

SPEAKER_01

But the fact that Google is iterating this quickly on their model lineup suggests they're feeling competitive pressure from anthropic and open AI.

SPEAKER_00

The flash branding is interesting. That usually implies speed optimizations, which makes sense if they're trying to compete on performance and cost efficiency.

SPEAKER_01

Right, and speed matters a lot for real-time applications. If Gemini Flash can deliver GPT level quality at significantly lower latency, that's a real competitive advantage.

SPEAKER_00

Twelve major announcements also suggests that Google is trying to demonstrate breadth across their AI portfolio. They're not just focused on one model, they're building out an entire ecosystem.

SPEAKER_01

That's classic Google strategy. Like throw a lot of things at the wall and see what sticks. But in AI, I wonder if that breadth approach is less effective than the focused approach we're seeing from Anthropic.

SPEAKER_00

It'll be interesting to see how the Gemini Omni updates compare to what Anthropic is doing with dynamic workflows. Both companies seem to be moving beyond single model solutions.

SPEAKER_01

Just a few months ago, Google felt like the clear leader. Now they're racing to keep up with startups that didn't even exist five years ago.

SPEAKER_00

Alright, let's step back and look at the bigger picture here. If you zoom out and look at everything we covered today, Anthropic's near trillion dollar valuation, their focus on model honesty, Apple potentially entering the AI assistant race, and enterprise companies getting more cost conscious about AI spending. What pattern emerges?

SPEAKER_01

I think we're watching the AI market transition from the Wild West phase to something that looks more like a traditional enterprise software market. The focus is shifting from wow, look what AI can do to how do we build sustainable, profitable businesses around AI that deliver real value.

SPEAKER_00

And that explains why Anthropic is emphasizing honesty and reliability over just raw performance gains. Enterprise customers care more about consistent, predictable results than they do about occasional moments of brilliance.

SPEAKER_01

Exactly. And it explains why Glean is succeeding with a cost-cutting message. Companies are moving from the experimentation phase to the optimization phase. They want AI that makes their business better, not just AI that's technically impressive.

SPEAKER_00

There's also this interesting tension between the massive valuations we're seeing and the increasing focus on cost efficiency. How do you reconcile a $965 billion valuation with a market that's becoming more price sensitive?

SPEAKER_01

I think it comes down to market segmentation. The companies that can demonstrate clear business value and sticky customer relationships will command premium valuations. But there's going to be a lot of AI startups that don't make it through this transition to a more mature market.

SPEAKER_00

The question is whether these trillion-dollar valuations can survive that transition. Mature markets typically have lower multiples than growth markets.

SPEAKER_01

The way Microsoft and Apple are platforms, then maybe these valuations make sense. But if they end up being just really expensive software tools, there's going to be a reckoning.

SPEAKER_00

And Apple's potential entry into the AI assistant market adds another wrinkle. If Apple can deliver good enough AI capabilities for free as part of their ecosystem, that fundamentally changes the competitive landscape for paid AI services.

SPEAKER_01

Right, which is why I think companies like Anthropic are smart to focus on enterprise customers rather than consumers. It's much harder for Apple to disrupt enterprise AI workflows than consumer application.

SPEAKER_00

The enterprise focus also makes sense from a business model perspective. Enterprise customers are willing to pay for reliability, integration, and support in ways that consumers typically aren't. That's how you justify billion-dollar valuations.

SPEAKER_01

And the dynamic workflows approach that Anthropic is taking feels like a bet that the future of enterprise AI is about orchestration and coordination rather than just having the smartest individual model. That's a platform play.

SPEAKER_00

So if you're trying to predict where this all goes, what are the key factors to watch? What will determine whether these valuations hold up and which companies actually succeed long term?

SPEAKER_01

Customer retention and expansion revenue will be crucial. Can these AI companies demonstrate that customers stick around and spend more over time? And can they build genuine competitive moats, whether through data integration, or specialized capabilities?

SPEAKER_00

The other big factor is how quickly the technology commoditizes. If running state-of-the-art AI models becomes cheap and easy, then the value shifts to application layer and user experience. That's where companies like Apple could have advantages.

SPEAKER_01

And regulatory factors could play a huge role. If governments start treating AI models like critical infrastructure or utilities, that could completely change the competitive dynamics and business models.

SPEAKER_00

One thing that's clear is that we're still in the early stages of figuring out what sustainable AI businesses look like. The companies that can navigate this transition from hype-driven to value-driven markets are going to be the ones that survive and thrive. That's our show for today. Lot to think about as the AI market continues to evolve and mature. What do you think about these trillion dollar valuations, sustainable or bubble? Let us know what you're seeing in your corner of the AI world.

SPEAKER_01

And if you're getting value from these daily conversations, make sure to subscribe so you don't miss an episode. The AI landscape changes so fast that yesterday's news is already old news.

SPEAKER_00

We'll be back tomorrow with more AI news and analysis. Until then, I'm Alex Shannon.

SPEAKER_01

And I'm Sam Hinton. See you tomorrow on Build by AI.