PLUS: AI designs new drugs, Apple's AI chief exits, and Runway's new video model
Good morning
NVIDIA is making a massive $2 billion investment into chip-design software company Synopsys. The strategic partnership aims to accelerate complex engineering workflows by moving them onto NVIDIA's powerful GPUs.
By shifting intensive design tasks from CPUs to GPUs, the collaboration could cut simulation times from weeks to just hours. Is this the move that solidifies NVIDIA's hardware as the essential foundation for the entire industrial design and engineering sector?
In today’s Next in AI:
NVIDIA's $2B bet on AI engineering
AI designs new drugs from scratch
Apple's AI chief steps down
Runway’s new model tops video benchmarks
NVIDIA's $2B Engineering Bet
Next in AI: NVIDIA is investing $2 billion in chip-design software giant Synopsys, forming a strategic partnership to accelerate complex engineering workflows with AI and GPUs.
Decoded:
The collaboration aims to move compute-intensive design tasks from traditional CPUs to NVIDIA's GPUs, slashing simulation times for complex products like computer chips from weeks down to hours.
Key initiatives include integrating NVIDIA's technology to advance agentic AI for autonomous design, building digital twins with NVIDIA Omniverse, and optimizing Synopsys' entire portfolio of applications.
The $2 billion investment gives NVIDIA a 2.6% stake in Synopsys, solidifying a non-exclusive partnership that still allows Synopsys to work with competitors like AMD and Intel.
Why It Matters: This partnership signals a major push to embed AI-acceleration deep into the core tools used by engineers in high-stakes fields like aerospace and automotive. It demonstrates NVIDIA's strategy of using its massive capital to ensure its GPUs become the foundational platform for the next wave of industrial design.
AI's Drug Design Revolution
Next in AI: Researchers have successfully used an AI model to design entirely new antibodies from scratch with atomic-level precision. The breakthrough, detailed in a new study in Nature, marks a significant advance in creating targeted treatments for diseases like cancer and influenza.
Decoded:
The AI model, RFdiffusion, overcomes a major historical challenge by accurately constructing the complex antibody loops that bind to therapeutic targets.
This research is directly tied to the commercial world, as key authors are co-founders of Xaira Therapeutics, a startup that launched with an eye-popping $1 billion in funding to advance AI-driven drug discovery.
This de novo design approach allows scientists to fine-tune therapeutic properties on a computer, potentially unlocking new ways to target hard to drug diseases and dramatically speeding up discovery timelines.
Why It Matters: This shifts antibody creation from a process of slow experimental discovery to precise digital engineering. While these AI-designed proteins are not yet drugs, the work establishes a powerful framework for augmenting how we create future medicines.
Apple's AI Shakeup
Next in AI: Apple's AI and Machine Learning chief, John Giannandrea, is retiring following significant delays in the rollout of the company's revamped Siri and 'Apple Intelligence' features, Apple announced yesterday.
Decoded:
Former Microsoft and Google AI researcher Amar Subramanya will take over, reporting directly to engineering chief Craig Federighi instead of the CEO.
The change follows Apple’s failure to deliver on major Siri upgrades promised for iOS 18, with key features now delayed until 2026.
This leadership shift comes as experts and investors believe Apple has fallen behind rivals like Google and OpenAI in the generative AI race.
Why It Matters: This shakeup shows Apple is making a decisive move to fix its AI strategy after public product stumbles. Placing the division under its seasoned engineering leadership signals a new focus on execution to catch up with competitors.
Runway's Video Upset
Next in AI: AI startup Runway launched Gen 4.5, a new text-to-video model that has already surpassed competitors from Google and OpenAI on a key independent benchmark. The new model excels at generating high-definition video from text prompts by better understanding physics and motion.
Decoded:
CEO Cristóbal Valenzuela emphasized the achievement, noting his team of 100 people out-competed trillion-dollar companies, internally codenaming the project 'David' in a nod to the biblical story.
Gen 4.5's top spot comes from blind comparison tests where human voters preferred its output over Google's Veo 3 (No. 2) and OpenAI's Sora 2 Pro (No. 7).
The model is rolling out gradually and will be available to all Runway customers by the end of the week through its platform, API, and select partners.
Why It Matters: This proves that smaller, focused teams can still lead innovation in the capital-intensive AI space. For creators and developers, this means more powerful and accessible video generation tools are becoming available outside the largest tech ecosystems.
AI Pulse
DeepSeek unveiled DeepSeekMath-V2, a new model focused on self-verifiable mathematical reasoning that scored a near-perfect 118/120 on the Putnam 2024 competition.
Telegram launched Cocoon, a new decentralized AI compute network on the TON blockchain, allowing users to rent out idle GPUs for AI queries in exchange for Toncoin.
Cloudflare announced it is acquiring Replicate, the AI deployment platform behind the popular Cog packaging tool, to integrate model inference directly into its developer stack.
Flock uses overseas gig workers to annotate sensitive footage from its U.S. surveillance camera network to train its AI models, according to an investigation of accidentally exposed company training materials.