In 2025: the great Chips Race, Video AI and more?
Last week: Google introduces quantum chip Willow, Apple and OpenAI cozy up
As we turn the corner into 2025 — here are some early signals and predictions we’re seeing on the horizon for AI and the future of creativity and computing.
What we got right in 2023 -2024
Voice as AI Interface
We wrote about the Future of Voice AI in Sep 2023, and 2024 has been the year of voice agents and interfaces.
Physical AI
We saw multimodal AI take off late 2023 with continued rigor in 2024, with significant capital going into physical AI companies like World Labs, Physical Intelligence, Archetype, and more. This is likely to be a multi-year journey, but we’re making interesting strides to push the research frontiers from varied approaches.
AI hardware
Bear-ish on AI hardware, despite the hype in early 2024. Likely to continue through 2025.
Signals we are looking closely at 2024-2025
Video AI
80% of internet traffic is video, and it's the fastest-growing communication medium for both people and businesses. I think we're finally at the inflection point for Video AI – from groundbreaking generative models to the infrastructure backbone supporting them.
The Great Chips Race
Apple, Google, and other tech giants are diving into the great Chips Race as AI demands surge. Just look at Nvidia and TSMC's stunning market performance (Nvidia's 5-year return of 2,307.21% still blows my mind) – and the demand keeps climbing.
Confidential Computing
As enterprise adoption grows and AI agents become more prevalent, security and privacy needs are becoming paramount.The thing about security companies – it's a winner-takes-most game, and the most in-demand providers tend to dominate.
Software Business Model May See Drastic Change
Users will be less interested in generalized software subscriptions, more interested in personalized software custom-built for them or their business. Powered by custom models and training datasets, and a personalized interface. They may be willing to pay more.
The Super Stack
The speed to go from idea to production is shrinking dramatically, and the race to become essential in builders' toolkits – from coding agents to infrastructure providers – is intensifying.
Lean Giants
We're witnessing the rise of what I call "Lean Giants" – startups scaling from zero to millions in ARR within weeks, running on surprisingly small teams. What once required 30 people can now be accomplished by just 3. I expect we'll see more of these lean operations making outsized impacts.
And with that, we wrap up our final newsletter of 2024. Happy holidays and see you in 2025!
The Latest This / Last Week
Google introduced Willow, a state of the art quantum computing chip with a breakthrough — cracking a 30-year challenge in the field. In benchmark tests, Willow solved a standard computation in <5 mins that would take a leading supercomputer over 10^25 years, far beyond the age of the universe. Google CEO Sundar Pichai says he sees Willow as an important step to build practical applications in areas like drug discovery, fusion energy, battery design.
Gemini 2.0 and Project Astra make Google's AI your know-it-all assistant: Google is entering its AI agents era with the introduction of Gemini 2.0 – the company's next generation AI chatbot – and a limited release of Project Astra, a computer vision-assisted AI agent that can see and analyze the world around you. Accompanying Gemini 2.0 are experimental “agentic” prototypes built to explore the future of human-AI collaboration, including:
Project Mariner, a prototype Chrome extension that can help with more complex tasks. Currently limited to Chrome for trusted testers, it can analyze text, images, graphs and other web elements at the pixel level and use that information to complete complex tasks.
Jules, an AI-powered assistant built for developers, that integrates directly into GitHub workflows to address coding challenges. It can autonomously propose solutions, generate plans, and execute code-based tasks – all under human supervision.
Google DeepMind is working with gaming partners like Supercell on intelligent game agents. These experimental AI companions can interpret game actions in real-time, and suggest strategies.
OpenAI has finally released Sora: OpenAI launched Sora, its text-to-video AI model. OpenAI says you can generate up to 50 priority videos (1,000 credits) at resolutions up to 720p with 5-second durations. The company highlighted a feature called “storyboards” that let you generate videos based on a sequence of prompts, as well as the ability to turn photos into videos. OpenAI also demonstrated a “remix” tool that lets you tweak Sora’s output with a text prompt, along with a way to “blend” two scenes together with AI. OpenAI says videos generated with Sora will have visible watermarks and C2PA metadata to indicate they’re made with AI.
OpenAI rolls out ChatGPT for iPhone in landmark AI integration with Apple: OpenAI demonstrated its new iPhone integration as iOS 18.2 rolled out to users, bringing ChatGPT directly into Siri, writing tools and camera features.
Meta AI approaches 600 million monthly users with Llama 3.3: Meta has announced that its AI assistant, Meta AI, is approaching a significant milestone with nearly 600 million monthly users. Llama 3.3 introduces a new 70 billion parameter model that Meta claims mirrors the performance of its previous 405 billion parameter model while being more cost-effective to run.
Optical chip interconnect startup Ayar raises $155M to bring light to AI workloads: Ayar Labs Inc., a startup that’s developing optical interconnects that can transmit data between computer chips via light, has closed on a hefty late-stage $155 million funding round led by Advent Global Opportunities and Light Street Capital.
Google’s new Trillium AI chip delivers 4x speed: Google unveiled Trillium, its sixth-generation artificial intelligence accelerator chip, claiming performance improvements that could fundamentally alter the economics of AI development while pushing the boundaries of what’s possible in machine learning. The custom processor, which powered the training of Google’s newly announced Gemini 2.0 AI model, delivers delivers a 4.7x increase in peak compute performance per chip compared to its predecessor, while doubling both high-bandwidth memory capacity and interchip interconnect bandwidth. Google has connected more than 100,000 Trillium chips in a single network fabric, creating what amounts to one of the world’s most powerful AI supercomputers.
Apple reportedly developing AI server chip with Broadcom: Apple is working with semiconductor company Broadcom on its first server chip designed to handle AI applications, according to The Information, which cited three people with knowledge of the project. Apple is known for designing its own chips – called Apple Silicon and primarily manufactured by TSMC – for its devices.
Midjourney launched Patchwork, a web-based tool for collaborative world-building and storytelling. This standalone application leverages multiple large language models to generate images and text, facilitating the creation of intricate narratives and interactive worlds.
Cognition Labs launched Devin, an AI assistant for engineering teams, for $500/month. The AI assistant can handle tasks like frontend bug fixes, backlog PR creation, and codebase refactoring, allowing engineers to focus on higher-priority work.