How the AI Price Wars Will Impact You

Today on the AI Daily Brief, the AI price wars heat up and we're talking about what it means for you. Before that in the headlines, Apple confirms its AI strategy is a complete mess The AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show. Notes. One of the substories we've been following for some time is just what the heck is going on with Apple when it comes to AI. This is a company that you really would think would be excellently positioned to understand exactly how to bring some version of AI to consumers in a way that actually met their needs, didn't require them to buy into a bunch of AI hype, and yet they have just done absolutely nothing. Underwhelming concept after underwhelming concept, followed by delay after delay. And now its Siri crisis seems to be deepening as a leaked all hands meeting speaks to disarray in the AI division. Now, of course, at this point, the troubles behind the AI versions of Siri have been well reported, including on this show, but comments from the company have been relatively slim. Last week, Apple confirmed for the first time that Siri was facing major delays and seemed to tacitly acknowledge that they've been advertising features that don't exist. Remember, they pulled ads from YouTube? Bloomberg's Apple Insider Mark Gurman reports that Apple Senior Director Robbie Walker recently addressed the team. Gurman wrote Walker suggested that employees on his team may be feeling angry, disappointed, burnt out and embarrassed after the features were postponed. The company had been racing to get the technology ready for this spring, but now the features aren't expected until next year at the earliest. You'll remember that AI Siri was first shown off at the Worldwide Developer Conference last June. Bloomberg has reported that at the time there was barely a working prototype and the unveiling relied entirely on a video mockup. Walker said that the delays were especially ugly and embarrassing due to the advertising push, stating this was not one of those situations where we get to show people our plan after it's done. We showed people before. He acknowledged that the marketing department had, quote, made matters worse by wanting to promote the enhancements even though they weren't ready, and touching on the true state of the delays. Walker even raised doubt about having AI Siri ready to ship in next year's iOS 19. He said that even though the feature is earmarked for release in the new version of the operating system, it quote, doesn't mean we're shipping then confirming what we already knew. Walker stated that the delays are due to quality issues with Siri's new features, quote, that resulted in them not working properly up to a third of the time, citing new hardware and software initiatives the AI team is working on. Simultaneously, Walter commented that we have other commitments across Apple to other projects. We want to keep our commitments to those, and we understand those are now potentially more timeline urgent than the features that have been deferred. He said the decisions will be made on a case by case basis, implying that AI Siri is not the number one priority. Finally touching on responsibility for the botched rollout, Walker insisted that there was, quote, intense personal accountability shared by his boss John Giannandria, the head of AI at Apple, as well as software chief Craig Federighi and other executives, gurman wrote as a Friday, Apple doesn't plan to immediately fire any top executives over the AI crisis, according to people with knowledge of the matter. That decision could theoretically change at any time. In any case, the company is poised to make management adjustments. It has discussed moving more senior executives under Giannandre to assist with the turnaround effort in a sign of just how far Apple has fallen. The leading Apple blog for the last 20 years at this point, Daring Fireball published on Wednesday a piece called Something is Rotten in the State of Cupertino. He said, in the two decades I've been on this racket, I've never been angrier at myself for missing a story than I am about Apple's announcement on Friday that the more personalized Siri features of Apple Intelligence scheduled to appear between now and WWDC would be delayed until the coming year. I should have had my head examined. Now Jon Gruber, the author of this Is raising Red Flags that something has gone deeply wrong at the company. He referenced a famous meeting from 2008, when Steve Jobs berated engineers on why an early iPhone email client doesn't work properly, direct quotes from Jobs included. You've tarnished Apple's reputation and you should hate each other for having let each other down, gruber wrote. Tim Cook should have already held a meeting like that to address and rectify the Siri and Apple intelligence debacle. If such a meeting hasn't yet occurred or doesn't happen soon, then I fear that's all, she wrote. The ride is over. When mediocrity, excuses and BS take root, they take over. A culture of excellence, accountability and integrity cannot abide the acceptance of any of those things and will quickly collapse upon itself with the acceptance of all three. I've been reading Gruber's blog for over a decade and always saw him defend every obvious Apple flaw. A true unapologetic fanboy. Seeing this is a complete shock. If even this doesn't prompt drastic leadership changes, I don't know what will. Meanwhile, in another big tech house, Google is taking the plunge in replacing the Google Assistant with AI. On Friday, Google announced that over the coming months, mobile users will be switched over to Gemini Assistant and the classic Assistant will no longer be accessible, they added. Additionally, we'll be upgrading tablets, cars and devices that connect to your phone, such as headphones and watches to Gemini. We're also bringing a new experience powered by Gemini to home devices like speakers, displays and TVs. Google notes they're continuing to focus on improving the quality of the day to day Gemini experience, especially for those who have come to rely on Google Assistant now. In some ways this was inevitable, but it's still noteworthy relative to the other companies in the space. Neither Amazon or Apple have even shipped an AI driven version of their assistant, let alone have the confidence to remove a legacy product from customers. David Barnard writes, as a recovering Apple apologist, I get the Apple has high standards and won't ship LLM Siri until it's ready argument, but I gotta call BS on that. Siri is a terrible product that we've all been complaining about for over a decade. The bar is low. Google is making the right choice here. Pushing to the future. Even though Gemini isn't perfect, it will get better quickly with hundreds of millions of people using it daily. And for a bunch of things it's already quite good. Would be funny to see Googling into how far ahead they are with Imamax style ads making fun of how dumb Siri is compared to Gemini. So there you have it. It's a mess in Cupertino, maybe a little bit better in Mountain View, and all of us are just watching it play out. For now though, that's going to do it for the AI Daily Brief Headlines edition. Next up, the main episode. Today's episode is brought to you by Vanta. Trust isn't just earned, it's demanded. Whether you're a startup founder navigating your first audit or a seasoned security professional scaling your GRC program, proving your commitment to security has never been more critical or more complex. That's where Vanta comes in. Businesses use Vanta to establish trust by automating compliance needs across over 35 frameworks like SoC2 and ISO 27001, centralized security workflows, complete questionnaires up to 5x faster and proactively manage vendor risk. Vanta can help you start or scale up your security program by connecting you with auditors and experts to conduct your audit and set up your security program quickly. Plus, with automation and AI throughout the platform, Vanta gives you time back so you can focus on building your company. Join over 9,000 global companies like Atlassian, Quora and Factory who use Vanta to manage risk improve security in real time For a limited time, this audience gets $1,000 off vanta@vanta.com NLW that's V A N T A.com NLW for $1,000 off we talk a lot about agents on this show, but if you've ever thought to yourself, I don't want to talk about agents anymore. I just want to actually build and deploy something. I'm really excited to share something special with you today. We've partnered with Linde to offer companies that just want to dive into the deep end of agents a way to get their feet wet, a way to move fast and build something meaningful without breaking the budget. The first five companies that email me NLWSUPER AI with Lindy in the title will have access to work with Lindy to build an actual functional agent serving their specific needs for under $20,000. Some of the agents you can build include a customer support agent, maybe automating responses on your website. You could build an SDR for generating or qualifying sales leads, or you could build an agent that's perfectly suited for your internal communications needs, be it note taking, scheduling, or something else. Not only is Lynde structured to integrate with all of the places that you already keep data and information, it's also a full, extensible platform, which means as you hire more and more agent employees and really build out your digital workforce, Lindy is going to enable those agents to be interoperable and basically be able to work together in a seamless way. So again, if you are interested in diving in all the way to agents in a matter of weeks, not months, not years, email me NLWSUPER AI. Put Lindy in the title and let's get your first digital employee online. Hey listeners, are you tasked with the safe deployment and use of trustworthy AI? KPMG has a first of its kind AI Risk and Controls Guide, which provides a structured approach for organizations to begin identifying AI risks and and design controls to mitigate threats. What makes KPMG's AI risks and controls Guide different is that it outlines practical control considerations to help businesses manage risks and accelerate value to learn more go to www.kpmg.usaiguide. that's www.kpmg.us aiguide. Today's topic is one that on first glance feels very de rigueur for the moment we're in Specifically, China's Baidu has released two new AI models which as always, they claim to have performance as high as, or at least close to the big American models, specifically GPT, and to do so for cheap. Except the cheap in this case is really cheap. We're talking about a claim that Baidu's new Ernie model matches or exceeds the performance of GPT 4.5 at about 1% of the price, $0.55 per million input tokens, as opposed to $75 per million input tokens for GPT 4.5. The other model that Baidu released, Ernie X1, which is their reasoning model, is priced at even 50% of Deep Seq's already low price. And so the sense that many people have is that this is a serious ratcheting up of the AI price war. Today we're going to talk about how this has been evolving and what it potentially means for a number of different groups. Now there is a sense, broadly speaking, that the price of AI was going to come down precipitously. In fact, even holding aside the onslaught of Chinese models, intelligence has been getting cheaper at a rate that far exceeds, for example, Moore's Law, which was the previous way that the technology industry thought about the speed at which technology became less expensive. There's a phrase that Sam Altman has been fond of intelligence too cheap to meter. In July, when OpenAI introduced GPT 4.0 Mini, he pointed out that just two years earlier, the best model in the world was not only much, much worse than the current models, but also 100 times as much. And of course, we've also seen, even within the big tech companies, price as a major competitive feature. Amazon, which still hasn't exactly gotten its feet under it when it comes to its own proprietary models, introduced its Nova foundation family in December, and it was very clear that part of the strategy was to compete on price. Google has also been trying to use price as a competitive advantage. In February, when Google released its Gemini 2.0 flash and flashlight, again, the major highlight was how much less expensive they were. Indeed, broadly speaking, the price of LLMs and the intelligence they represent has been just absolutely collapsing. But of course, all of this took on a new dimension when Deepseek launched, claiming that their model, which had very similar performance over comparative OpenAI models, was trained for less than $6 million. That revelation, unconfirmed, as though it may be, has rocked the markets and hasn't really let them go. When Deepseek R1 came out, people started calling it the Sputnik moment, launching a global race for ever cheaper AI. So profound was the psychological mark that Deepseek left in terms of recalibrating how people thought about where China was in the AI competition, that everything subsequent to that has been Is this the next Deep Seat moment? Last week we talked about the AI agent Manus, which many people called China's second Deep Seat moment. And once again, while Manus hadn't innovated at the foundation model level, it had created a consumer product that just seemed to beat everything that was available here in the us. Now, once again we have people calling this Baidu Drop another Deep SEQ moment. Baidu writes, We've just unveiled Ernie 4.5 in X1 as a deep thinking reasoning model with multimodal capabilities. Ernie X1 delivers performance on par with Deepseek R1 at only half the price. Meanwhile, Ernie 4.5 is our latest foundation model and new generation native multimodal model. Now, in terms of capabilities, these models have everything you'd expect. They can analyze and summarize documents, they can solve complex problems, but price is really what people are talking about. Now some have pointed out that while Ernie's X1 model is about half the cost of Deepseek's R1 reasoning model, Deepseek's V3 non reasoning model is still about half as much as Ernie 4.5, although both of those are of course dramatically cheaper than both GPT 4.0 and GPT 4.5. In response to this, the memes were flying quickly with a video of a multi car pileup. Jeffrey Townsend writes a real time view of investors in Genai foundation models Low cost deepseek shocked OpenAI now baidu has released super low cost journey 4.5. China is driving the cost of AI way down. It's brutal. Sandeep Mahaden writes, China's AI firms are not only building fundamentally better models, for example deep seq, they're building fundamentally cheaper models. America cannot compete with this if it continues. This is radical innovation, state funded or otherwise. So what really are the implications of this? Well, first of all there's the stock market and obviously we've seen that the emergence of these models and the idea that they might use much less compute to be able to get this sort of performance threatens the narrative of companies like Nvidia, which have driven the rally For a couple of years now. I think it's important to have a couple of caveats here. First of all, there's a lot more going on than deep seek when it comes to stock market woes. Right now we're in a period of extreme volatility and unpredictable futures and markets are not just dealing with China and AI, but also with tariffs, geopolitical realignment. So trying to parse out how much tech stock underperformance has to do with that versus just a correction after two years of basically unfettered up into the right is a little bit harder to parse than people might be making it seem. The other question, of course more structurally when it comes to compute is is cost of inference. The Wall street narrative is still kind of stuck on the idea that the only use of compute is to train new models rather than to deliver those models in practice. The counter argument, the one that companies like Nvidia have been making, is that the cheaper the models become, the more people use them. The more people use the models, the more inference costs they incur. And so the burden of the compute shifts to a different part of the stack, but still remains regardless. It makes things look like a less clear bet and that could have implications for downstream funding as well. Now, when it comes to startup business models, it's a mixed bag. In the short term, there's a lot that's amazing about this. Downward price pressure means that all the startups out there can do and offer a lot more for a lot cheaper. The more intelligence becomes available in a cost effective way, the more startups are going to find ways to use it. And that's a very good thing. In the long term, it could be a little bit more challenging. If the price compression continues to be as severe as it looks, it could constrain and limit the band of prices that startups can actually offer. One place that this will specifically come to bear is with regard to agents. Right now there are lots of different pricing models. When it comes to agents. People are experimenting with outcome based pricing and generally trying to think about things in new ways outside of the traditional SaaS model. But I would say that by and large, they're still benchmarking it against the comparative human labor. If you have a sales or SDR agent, the promise of that agent is that you're going to pay less than the equivalent human time would have cost. However, in terms of how much less, companies are still benchmarking it against the human that would have done the job before. And that still makes them pretty expensive. It seems highly likely to me that someone is going to try to reverse this flow, and instead of pricing it on the basis of what the comparative human time would have been, they're going to price it on the basis of the cost of goods and have a radically cheaper price. That undercuts the entire premise of that other model. Now, once again, the countervailing pressure here is that if I'm right and in the future, we don't just hire one agent to do the thing that a person used to do, but a thousand agents to do it in a totally different type of way, maybe that all ends up in a wash, but still, the point is that the price war will have impacts on both the startup side in terms of what they can offer, as well as the enterprise side in terms of what they expect to buy. And then of course, there's the geopolitical dimension of all of this. One of the questions is how much is this intentional price warfare? Is this China and Chinese companies doing something unsustainable, in fact and uneconomical, in order to cause harm to American competitors? The answer is probably that it doesn't matter as long as the companies engaging in the price war have deep enough pockets to keep it going. Certainly the American companies aren't loving it. In OpenAI's proposal for the USAI action plan, they basically argue that deep SEQ and Chinese AI should be banned, which of course is one way to limit competition. There are other strategies though. Robert Scoble, for example, wrote over the weekend, if I were Mark Zuckerberg, I'd release a badass AI model for free and end this price war once and for all. Why? Because the model that wins will collect more real time data from all of its 3 billion users, which will make its glasses and services better and more profitable. And indeed, there is a sense that perhaps this is the path Mark Jeffrey writes AI smartness goes to infinity AI price goes to free open source style vendors of add on services and verticals built around AI win Embodiment of AI in the world of atoms wins and becomes geometrically more valuable with intelligence increases There is a lot up in the air right now when it comes to the future of the business model of AI. At this point all that seems clear is that the business side of this is going to change nearly as fast as the technology side. I hope if I've convinced you of anything today, it's that there are actually meaningful implications of this to the AI you interact with, what price you pay for it and the opportunities that it creates for you and so I will, of course, keep track of how these things evolve and change over time. For now, though, that's going to do it for today's AI Daily brief. Until next time, Peace.