Everyone from OpenAI to SpaceX is building their own chips | Equity Podcast

Custom Chip Development and Strategic Motivations

• OpenAI is building a custom chip called Jalapeno with Broadcom to reduce costs, with a focus on inference, and this move may provide OpenAI with an alternative to relying on Nvidia GPUs 10s.
• The development of custom chips is a trend among companies, with Google working on its own designs for custom hardware to aid in AI development, and SpaceX planning to work with Tesla to develop a Terafab 2m6s.
• The benefits of developing custom chips include hedging against single supplier risk and being able to customize the hardware to meet specific needs, with many companies willing to spend resources on this effort 2m6s.
• Sean O'Kane, now a senior reporter for special projects, is looking to collaborate with others in the newsroom on big projects with significant impact, and is open to receiving tips and information from the public, as long as it is not embargoed 42s.
• The discussion on chips is one of the main themes for the week, with several related headlines and companies working on their own custom chip designs, including OpenAI's Jalapeno and SpaceX's planned Terafab 10s.
• Apple's decision to break with Intel and design its own in-house silicon has led to numerous benefits, including making its products more desirable due to the custom silicon, and this concept is also being applied by other companies like OpenAI, with the added benefit of distributed risk 10s.
• The goal of creating custom chips is not only about creating distance in a complicated landscape where companies both work with and compete with each other, but also about lowering costs, particularly for certain tasks like inference-related tasks and coding-related AI 2m6s.
• Lowering costs is a crucial aspect, as the current demand for memory chips is skyrocketing, with very little supply, leading to price increases, and companies like OpenAI are looking to reduce costs at any point in the process to mitigate this issue 2m6s.

Market Impacts and Consumer Consequences of Chip Shortages

• The memory chip crunch is affecting consumers, with companies like Apple raising their prices due to demand problems, and it is likely that consumers using AI services like ChatGPT will also see price increases 4m30s.
• If companies like OpenAI are able to produce chips that can perform tasks cheaper, the question remains whether these savings will be passed on to consumers or used to help the company get closer to profitability, especially with OpenAI potentially going public later this year 6m40s.

Nvidia's Dominance and Strategic Moves in the AI Chip Market

• The stock price of SpaceX has been volatile since it went public almost two weeks ago, and it is expected to have a harder time on the markets, with a potentially more volatile time ahead 10s.
• The announcement of a new chip project mentioned lowering costs and making things better, but more information is needed to understand the progress made, and numbers are expected to be seen in the S-1 filing and subsequent filings as the company goes public 42s.
• Nvidia is a company that is still dominating and making progress in the field, either through acquiring companies, aqua-hiring companies, or doing their own research and development, with a recent example being their deal with Groq 2m6s.
• Groq, a company that was thought to be left for dead after Nvidia pulled out its leadership, has raised $650 million and seems to be doing well, which is a surprise and an example of a company that has found a way to succeed after a reverse acquisition 4m24s.
• Nvidia's success can be attributed to their pivot towards operating an AI cloud platform rather than competing primarily as a chip company, which has allowed them to manage to get a $650 million raise and continue to dominate the field 6m40s.

Reverse Acquisitions and the Rise of Startups in the Tech Ecosystem

• The trend of reverse acquisitions, where a company acquires the talent and technology of a startup but leaves the startup itself to continue operating, has been seen in several cases, including Groq and Scale AI, and it is interesting to see how these companies can still find a way to succeed and grow after such a deal 8m10s.
• Many companies, including Groq and SpaceX, are leasing out their compute resources, as numerous players are compute-constrained and see an opportunity to generate revenue by doing so 10s.

Space-Based Data Centers and Their Feasibility

• SpaceX is continuing to rent out its compute to other companies, having recently signed a deal with a smaller player, and is looking to build out this revenue stream, although the durability of this business model is uncertain 1m15s.
• Masayoshi Son, the CEO of SoftBank, has questioned the point of having data centers in space, highlighting the challenges and high costs associated with this approach, and noting that it will not provide a solution to immediate data center needs 2m6s.
• SoftBank has a history of making bold bets, and Son's questioning of orbital data centers is seen as a significant moment, as it highlights the need for scrutiny of this idea, which has gained significant traction in recent years 3m30s.
• The concept of putting data centers in space poses significant engineering and economic challenges, but companies like SpaceX, led by Elon Musk, are likely to pursue this idea due to the potential benefits of avoiding regulatory hurdles and environmental concerns 4m45s.
• SpaceX's launch business is currently heavily reliant on its Starlink project, which drives a significant portion of its launch market share, and the company's efforts to develop its AI and data center business are likely aimed at reducing this dependence 6m15s.

The Launch Market and Its Dependence on Starlink

• The launch market is dominated by a single company, with Starlink being a significant factor, and the need to replace satellites every few years guarantees more business for the launch industry 10s.
• Companies like SpaceX have other big businesses, such as renting out their compute, which is related to the conversation about chips and the development of new technologies 2m6s.

Tech Executives and the Art of Future Prediction

• Executives at tech companies often predict a future that is advantageous to their business, which is not a new phenomenon, and it is essential to remember this when discussing big AI companies and their predictions for the future 2m6s.
• The concept of an orbital data center, which involves a constellation of satellites, is being discussed by various companies, including SpaceX, and some notable figures, such as Sam Altman, have expressed skepticism about this idea 4m30s.

Emerging Trends in AI: Loops and AI Agents

• The term "loops" has become a new topic of discussion in the AI industry, referring to AI agents that run endlessly in the background, prompting other agents, and having significant implications for the field 8m30s.
• Loops are being talked about as a crucial development, potentially more important than source code, and some companies, such as Anthropic, are highlighting their significance, which may be seen as "talking their own book" 10m40s.
• The focus on loops may be a temporary trend, with the AI industry constantly evolving and new terms and concepts emerging, such as token maxing, which was a hot topic a few months ago 12m20s.
• The concept of AI agents always working may not live up to the hype surrounding it, and the excitement about this technology might be deflating and disappointing in the long run 10s.
• The idea of loops, which is the next iteration of the conversation about AI agents, may have significant implications, but its widespread use is uncertain, and there are reasons to be skeptical about its adoption 42s.

Enterprise Applications and Limitations of AI Agents

• Anthropic is focused on enterprise users, particularly those who use AI for programming, and its technology seems well-suited for these tasks, but it raises questions about the ability to control and understand the actions of AI systems 2m6s.
• The use of AI agents and loops may be limited to specific enterprise applications, and it is unclear whether regular consumers will have complex enough problems to require the continuous use of AI agents 2m6s.
• Current AI technologies, such as Google Gemini, Anthropic, or ChatGPT, can already resolve many complex tasks quickly, which may reduce the need for continuous AI agent use in everyday life 2m6s.

Challenges and Realism in AI Agent Adoption

• The concept of looping and its quality is an important aspect that has not been fully discussed, and it is unclear whether it will be broadly applied to all of enterprise or specific groups 10s.
• There is a notion that the entire back office of a company can be replaced, and possibly some front office components, which would necessitate the use of advanced technologies like looping, but this has not been achieved yet due to the brittleness of current agentic technology 2m6s.
• The idea of using looping for workplace purposes outside of programming is being considered, but it still lacks the robustness to be effectively implemented, and companies are now focusing on controlling the costs associated with AI usage tokens 2m6s.

Cost Management and Value Realization in AI Investments

• The conversation around AI usage tokens has evolved rapidly, with companies initially excited about token maxing but now struggling to keep costs under control, and this has led to a closer examination of the value provided by these technologies 4m30s.
• Companies are experimenting with looping and agentic AI, but they need to balance the potential benefits with the need to keep an eye on the bottom line, and there is a growing demand to demonstrate the actual value and productivity gains from these investments 6m15s.
• The situation is likened to a scenario where companies are being sold a repositioned product with savvy marketing, but ultimately, the technology needs to deliver more than just marketing promises to be successful 8m40s.

Agility Robotics and the Resurgence of SPACs

• Agility Robotics, a company that has been around since 2015 and spun out of Oregon State University, is seeking to go public through a SPAC, valued at about $2.5 billion, and has been working on a humanoid robot called Digit, which has already gained customers and is looking to expand 10s.
• The use of SPACs as a financial instrument had faded away after being popular in 2021, but is now seeing a resurgence, with some expressing skepticism about their effectiveness due to past experiences where companies that used SPACs did not achieve long-term sustainable success 2m6s.
• Agility Robotics is considered one of the more mature companies to merge with a SPAC, having already generated revenue and secured multi-year orders worth over $300 million, and is moving out of the pilot phase with nine customer sites, including working with major companies like Toyota 4m6s.
• The company's decision to use a SPAC to go public may be driven by the need for liquidity, and while some are cautiously optimistic about its potential for success, others are more skeptical due to the past failures of companies that used this route 6m6s.
• Unlike many previous SPACs, Agility Robotics has a real product and customers, and its progress is seen as more reasonable, but the outcome of its decision to go public through a SPAC remains to be seen 8m6s.

Challenges and Opportunities in Going Public via SPAC

• Companies that raise money through a SPAC need to leverage their capital to fulfill orders and make money, which can be difficult, and the financial terms of the investment are crucial to the company's success 10s.

AI in the Film Industry and Cultural Backlash

• A24, an indie movie studio, has taken an investment from Google DeepMind to develop new AI tools for filmmakers, which may seem surprising given A24's reputation as a cool and indie studio, but it is likely a move to scale up and stay competitive 2m6s.
• The investment in A24 by Google DeepMind may indicate that Hollywood is cracking the door open to AI, but it is not yet widely accepted, and there is still pushback from unions and creative teams who are cautious about the use of AI in filmmaking 2m6s.
• The concept of "selling out" may be experiencing a resurgence, particularly among younger people, as evidenced by the backlash against A24's investment from Google DeepMind, and this could be a sign of a new era of criticism against companies that prioritize profit over artistic integrity 4m30s.
• A24's decision to take an investment from Google DeepMind may be a strategic move to stay relevant and competitive in the film industry, and it is not necessarily a surprise given the studio's history of taking on private equity money and partnering with other companies 6m0s.
• The use of AI in filmmaking is still a topic of debate, and while some studios may be experimenting with AI, there is still a lot of pushback from the film community, and it remains to be seen how widely accepted AI will become in the industry 8m0s.

Closing and Future Outlook

• The show Equity will be back next week, indicating a regular schedule, and the audience is being thanked for listening 0s