U.S. Gov't Grants Intel $8.5B for AI Chip Production

Today on the AI Breakdown, we're talking about the big new deal between the US Government and Intel to bring chip manufacturing back to the us. Before that, on the Brief, a surprising number of Americans think AI could make better TVs and movies than humans. The AI breakdown is a daily podcast and video about the most important news and discussions in AI. Go to breakdown.network for more information about our YouTube, our Discord and our newsletter. Welcome back to the AI Breakdown Brief on all the AI headline news you need in around five minutes, Here are some numbers that could make Hollywood a little bit nervous Deloitte has just released their 18th annual digital media Trend Survey. One of the things they asked were people's perspectives on TV and movies. The good news for Hollywood is that 70% of US consumers say they would rather watch a TV show or movie written by a human than generative AI. The bad news is that 22% or over a fifth said that they thought that Gen could actually write TV shows and movies that were more interesting than those produced by humans. And those numbers were higher among younger generations, with 25% of Gen Z thinking that and 30% of millennials nearly a third thinking that Gen AI could do better than humans. Overall, 42% said that both generative AI and humans could produce entertaining content. There were also some numbers in there around current usage, with around quarter of those younger generations having used AI to create text a la ChatGPT, and around 18% having used AI to create images. In the wake of Sora, the conversation around AI's impact on Hollywood has definitely gotten nothing but louder. We we covered on this show, for example, Tyler Perry announcing that he was halting an $800 million expansion of his studio specifically because of what he saw with Sora. And in a recent Harris X poll, consumers were asked to watch eight videos. Half of them had been produced by humans and half of them were SORA demonstration videos. Basically, they had no ability to actually distinguish. For example, one AI generated video of waves crashing at Big Sur had 60% of respondents say that it was created by a human. Now, the scale and the scope of AI rot change perhaps makes our next bit of news entirely expected. And that is the announcement that Anthropic is teaming up with Accenture and aws to train 1400 Accenture engineers as specialists using Anthropic's models on AWS. Anthropic writes, Accenture's engineers will help organizations use their own data to fine tune Anthropics models on AWS to enhance performance for their use. Case and industry. Ralph Brooks writes Translation Accenture has relationships with Fortune 500 companies to sell consulting work. Those companies use AWS and that isn't going anywhere. Anthropic has high end models but dismal market share, so Anthropic will do fine tuning bootcamps to Accenture for a fee One of the really notable things about the AI adoption curve to me is how aggressively the big consulting firms like Accenture are getting into it. Those types of firms are not just passive recipients of corporate business, they are drivers in pitching businesses on what type of change they need to undergo. The fact that they're making such a huge investment in this type of work is to me more evidence of just how on the precipice of AI rot change we really are when it comes to the enterprise. Patrick Hussey, however, points out something else about the deal. He writes something worrying is becoming clear in the AI transition. It's a game only giants can win and they are really going for the jugular. The three way deal between Anthropic, Amazon Web Services and Accenture highlights this new reality. Previous sea changes in business and tech were led by plucky startups that took years to grow. In this transition, that's not possible for two reasons. One, the training runs. Top foundation models require eat up chips, cash and compute. Small companies do not have to. The disruptive opportunities have already attracted established players who are pumping rocket capital and know how into only a few companies result. The new giants of AI are joining forces with legacy giants in consulting and tech to own enterprise AI adoption. This is a clear bid to seize new economic territory and own the future. This is something that I've covered a lot on this show, especially with things like the announcement this week that a big part of the inflection AI team was going over to Microsoft despite having raised 1.3 billion less than a year ago. Another piece that Patrick doesn't even mention, which I do think is important as well, is that because so much of the value of AI comes from how an enterprise lets it interact with their data, there is also a natural bias towards working with companies that they already trust to interact with their data, which gives incumbents the sales forces of the world, for example, a leg up that they might not have previously had in other tech revolutions. Another story reinforcing just how hard it is to be a foundation model startup if you're not at the very top level. The the Information reports that Cohere is having a slightly hard time raising funding. The Information writes bankers told investors late last year that Cohere was seeking new funds at a $6 billion valuation, which would be more than 450 times its annualized revenue of 13 million, according to two people who are familiar with the discussions. The Information points out that that is a much higher valuation multiple than we've seen for even other AI companies like, for example, OpenAI, which was valued at about 50 times its forward revenue. The information also writes that if Cohere is able to pull off this round, its valuation is likely to be a lot lower than that 6 million, although still leaving the company at a much higher valuation multiple than some other AI startups. So what is Cohere's strategy? The Information writes. Cohere has tried to differentiate itself from larger competitors by selling its technology only to enterprises and avoiding the consumer chatbot race sparked by ChatGPT. And in recent months, Cohere's leaders have decided not to compete with OpenAI and others in developing the biggest, most state of the art AI models, instead focusing on retrieval augmented generation. Neil Costa sums up what I think a lot of people are thinking looking at this deal when he writes, this technology will follow a power law. It's going to be hard if you're not in the top three or so. That is going to do it for today's AI Breakdown brief. Next up, the main AI breakdown Today's podcast is brought to you by Plum. You've probably noticed by now that many of the AI features that are embedded in your favorite products kind of suck. They're cool the first time, but pretty soon you're underwhelmed. That's because truly great AI features require complex pipelines and rigorous testing that most startups simply don't have time or tooling to get right. That's why Plum created a collaborative AI app builder that's purpose built for product teams. Your users deserve better than a glorified GPT wrapper. Blow their minds with plum. Check out useplum.com, that's plum with a B. Send me a note to get early access. Yesterday President Biden tweeted, since I took office, companies have announced over 240 billion in investments to bring chip manufacturing back home. That's the Chips act working. Today we'll announce a preliminary agreement with intel for up to 8.5 billion in chips funding mobilizing $100 billion so this story is about Intel a little bit, but broadly it's about the geopolitics of chip manufacturing in the context of an AI powered world where this is increasingly important. Now I'm not going to get too deep on this, but if you want to go learn about why the semiconductor industry isn't in the US currently, or more specifically how it left the US I would recommend checking out something like Perplexity. Just a quick search suggests the combination of cost of labor, government incentives, proximity to raw materials for the supply chain, and more as reasons and because Perplexity sources everything. You can go follow up whatever details are most interesting to you. Now the larger context for this intel deal is the Chips and Science act, which was signed back on August 9, 2022. The act authorizes $280 billion in new funding to bolster U.S. semiconductor research, manufacturing and development. Within that, 52.7 billion was earmarked for semiconductor manufacturing incentives and research initiatives, including 39 billion in subs for chip manufacturing facilities that are built on US soil, 13 billion for semiconductor research and workforce training, and a 25% investment tax credit for manufacturing equipment costs. So what exactly is this deal? Well, the first part of it is that the US government is giving Intel 8.5 billion in grants to bolster semiconductor production. This will subsidize the construction and expansion of intel facilities around the US and is so far the biggest funding that's come from the CHIPS act yet, said President Biden. Nearly all manufacturing of leading edge chips across the entire industry moved overseas to Asia years ago. That's why today's investment is such a big deal. We will enable advanced semiconductor manufacturing to make a comeback here in America. In addition to the 8.5 billion in grants, the government is also giving intel up to $11 billion in loans with what, as the New York Times put it, the company characterized as generous terms. So where will intel actually be investing? Well, first of all, the grants and loans are expected to catalyze an Overall total of $100 billion in private investment as well, and is expected to create around 10,000 manufacturing jobs and another 20,000 construction jobs. This will involve expanding operations in four states, including Oregon, Arizona, New Mexico and Ohio. And in some cases, these are some of the biggest investments those states have ever seen. In the case of Ohio, the White House claims it is the largest private sector investment in the state's history. One of the big things that's notable about this push to bring chip manufacturing back to the US Is it's not just a cost question. There is a major knowledge gap among American workers because this isn't something we've done for decades and decades, the White House writes. The announcement today also includes significant funding to train and develop the local workforce including 50 million in dedicated chips funding. Now, intel has gotten some serious press for its shifting efforts here. Back in February, Wired wrote a piece called Intel's AI Reboot is the Future of US Chip Making, and what the article discusses is a shift in strategy. Specifically, intel is transitioning to make a big bet on the foundry business. Their foundry business basically would have them work with other companies to manufacture those companies chips and Intel's ambitions are very big. CEO Pat Gelsinger said in February that generative AI is transforming everything about computing and through our foundry I want to manufacture every AI chip in the industry. That foundry business relaunch event back in February also featured Microsoft CEO Satya Nadella who said we will need a reliable supply of the most advanced, high performance and high quality semiconductors. That's why we're so excited to work with Intel Foundry Services now. As part of the big grant yesterday, Gelsinger actually published an op ed in Fortune magazine as well that displayed the full extent of their ambition. Our goal, he said, is to have at least 50% of the world's advanced semiconductors produced in the US and Europe by the end of the decade. In that piece, he writes, despite their unparalleled importance, America has allowed its control over semiconductor production to slip away with more than 80% of chips now manufactured in Asia. Having once pioneered the development of this extraordinary technology, we now find ourselves at the mercy of the most fragile global supply chain in the world. There will always be a global component to the chip industry, but Intel's moonshot goal is to have at least 50% of the world's advanced semiconductors produced in the United States and Europe by the end of the decade. After all, America invented the microchip. As we enter the AI era, which is poised to reshape the world, there's no reason why this country of thinkers and doers should not lead the the world in developing and manufacturing the most advanced chips. The CHIPS act combined with Intel's new and expanding semiconductor foundry sites from the Silicon desert of Arizona to the silicon heartland of Ohio, shows how we can once again do just that. It will be no mean feat to try to displace the current manufacturing giants like tsmc, although Timothy Green from the Motley fool does point out that the fact that Nvidia's new Blackwell system does not use TSMC's most advanced node suggests that TSMC is having some issues that could open up opportunities for companies like Intel. Whatever the case, by and large there is lots of excitement about this, summed up by Sam Altman who wrote happy to see this, excited for intel, the US and more AI capacity. That is going to do it for today's AI breakdown. Until next time, peace.