⚛️🦾 Quantum Cybersecurity, Neuromorphic Computing at Scale, and Project Stargate

A Newsletter for Entrepreneurs, Investors, and Computing Geeks

Author

Benjamin Wolba
January 27, 2025

This week, the West was captivated by the rapid AI progress in China, with DeepSeek’s latest reasoning model, R1, outperforming OpenAI’s o1. Also, DeepMind and Kimi announced new reasoning models, and there’s been a lot of buzz around Project Stargate and significant investments in AI infrastructure.

More players, like Neuronova, are exploring energy-efficient, neuromorphic chips to squeeze more intelligence out of watts. There’s also been exciting news around optical memory. And one startup, ZuriQ, managed to get its funding news published across all major tech outlets. Check out our interview with them last August.

Finally, a quick reminder about our Future of Quantum Computing Meetup on February 4 in Berlin. Also, don’t miss the Meta Llama Hackathon in Oslo on February 1-2.

Spotlights

Quantum Bridge, founded in 2019 by Mattia Montagna and Hoi-Kwong Lo, pioneers quantum-safe communications with a suite of products and services. Their Distributed Symmetric Key Establishment (DSKE) technology integrates classical cryptographic techniques with quantum-safe algorithms, safeguarding networks already today. At the same time, they’re developing quantum repeaters, the building blocks needed to build a secure and quantum-native internet.

“Deepseek is a Chinese AI startup whose latest R1 model beat OpenAI’s o1 on multiple reasoning benchmarks. Despite its low profile, Deepseek is the Chinese AI lab to watch.”

It was hard to miss if you were on X this week: DeepSeek just announced its reasoning model R1

It has committed to open-sourcing all its models and has significantly reduced inference costs through innovations like multi-head latent attention (MLA) and sparse mixture-of-experts (DeepSeekMoE). Price war is in full swing, not only in China.

Toronto-based quantum startup Xanadu has unveiled Aurora, a new photonic quantum computer that interconnects four modular server racks using 35 photonic chips and 13 kilometers of fiber optics, all operating at room temperature. This 12-qubit system represents a significant advancement in networking quantum computers.

🧠 Neuromorphic Computing at Scale (EENews Europe)

A recent review by leading European researchers, including Steve Furber from the University of Manchester and Hector Gonzalez of SpiNNcloud Systems, examines the roadmap for scaling neuromorphic computing to address the energy demands of artificial intelligence. Key strategies for achieving scalability include heterogeneous integration, event-based computation and communication, and efficient software capable of managing real-world complexities.

The U.S. government's recent decision to limit the export of AI chips to Poland has elicited strong objections from Polish officials, who argue that the move could hinder the nation's technological development and military expansion efforts. The U.S. has categorized countries into tiers regarding AI chip access, with Poland not included among the 18 nations granted unrestricted access.

The United States is preparing to expand trade sanctions to include legacy, or "mature node," semiconductors—essential components in everyday electronics like car remotes and refrigerators. This move aims to counter China's strategy of subsidizing these older chips to flood global markets, potentially undermining Western manufacturers.

Headlines

🦾 First-ever data center on the Moon set to launch next month (Techspot)

🦾 EU launches Chips JU pilot lines: a pivotal €3.7B initiative aimed at strengthening Europe’s semiconductor innovation and production capabilities (Innovation News Network)

🦾 ByteDance plans $20 billion capex in 2025, mostly on AI, sources say (Reuters)

🦾 Nvidia impact: World's second-biggest memory chipmaker from South Korea reports over 8 trillion won profit (MSN)

🦾 Low-Power Magnetoresistive Memories Are Here – follow Arkady Kulik for more great deep tech insight! (LinkedIn)

🤖 Introducing LFM-7B: Setting New Standards for Efficient Language Models (Liquid)

🤖 Introducing Kimi k1.5 --- an o1-level multi-modal model (Kimi.ai on X)

🤖 Gemini 2.0 Flash Thinking Experimental: DeepMind’s enhanced reasoning model (DeepMind)

🤖 When AI Passes This Test, Look Out: The creators of a new test called "Humanity's Last Exam” argue we may soon lose the ability to create tests hard enough for AI models (NYT)

⚛️ Myths around quantum computation before full fault tolerance: What no-go theorems rule out and what they don't (arXiv)

⚡️ Quantinuum to open photonics research center in New Mexico (Optics)

⚡️ Unlocking the Speed of Light: Researchers have unveiled a programmable photonic latch that speeds up data storage and processing in optical systems (SciTechDaily)

🧠 Bringing the brain to silicon: Neuronova's vision for energy-efficient AI hardware (Tech EU)

Funding News

⚛️ $4.2M Seed – ZuriQ: a fundamentally different approach to ion trap quantum computing (ZuriQ)

🦾 $36M Series B – Baya Systems: Accelerate Intelligent Compute with Chiplet Technology (Embedded)

🌐 $10B Series J – Databricks: cloud-based platform for storing and analyzing data (Silicon Angle)

Deep Dive: Project Stargate

It’s not yet $7 trillion, but $500B. This week, OpenAI announced Project Stargate, a new company formed by SoftBank, OpenAI, Oracle, and MGX that plans to invest this much money over the next four years to build AI infrastructure in the United States.

While there’s already been controversy about whether funding is secured, as  Elon Musk and Semianalysis pointed out, Reuters reported both OpenAI and Softbank have committed $19B already—that’s a start!

The initiative aims to secure American leadership in AI, create hundreds of thousands of jobs, and provide strategic capabilities for national security. Yet, much of the capital is funded by UAE sovereign wealth through MGX and SoftBank's high-risk investments. This leaves the question of how much America depends on external funding for critical technology sectors and the potential influence of these "shadow shareholders" on the nation's AI future.

The four-year timeline conveniently matches Trump’s presidential term, and the $500B figure, while impressive, seems more about optics than substance. The $100B starting point is likely more realistic. At least, OpenAI’s announcement leans heavily into patriotic rhetoric—perfectly in step with the MAGA narrative.

P.S. Here’s a cool blog post about what it takes to build a data center by folks who have done it: So You Want to Build Your Own Data Center (Railway Blog)

💡 Love these insights? Forward this newsletter to a friend or two who can subscribe here