Signal + Noise: GEN AI WEEKLY RUN DOWN, 1.29.25
By Isaiah Steinfeld, Founder & CEO, Neue Alchemy
January 29, 2025
Three weeks in, and the AI landscape isn’t just evolving—it’s taking shape in ways that are both exciting and uncertain. The foundation is solid, the infrastructure plays are in motion, and now? Now we’re watching AI move from hype to execution—but not without turbulence.
This week’s shift is clear: AI itself isn’t the story anymore. It’s what you do with it that matters. The market is rewarding startups solving real problems, while the biggest players are placing infrastructure bets that could either reshape the future—or just lock in control.
The signal? AI’s next phase belongs to the ones who can turn potential into something lasting, something meaningful. Not just flash—but follow-through.
Let’s dive in.
🚀 Stargate: The $19B Power Play
Last week’s $500B Stargate vision sounded ambitious. This week, OpenAI and SoftBank put down $19B each, signaling this isn’t just a concept—it’s a real attempt to reshape AI infrastructure at scale. One data center in Abilene TX is just the start.
💡 Why This Actually Matters:
• Infrastructure Isn’t the Bottleneck Anymore – With hyperscalers and sovereign AI initiatives taking shape, the focus shifts to who builds on top of it. As Arjun Dev Arora (Managing Partner at Format One) pointed out, the foundational layer is set—now it’s time for startups to execute, applying AI to solve real-world problems.
• Global Competition Is Heating Up – UAE’s MGX AI Fund is eyeing a stake, raising big questions about who really controls AI infrastructure.
• Not Everyone’s Convinced – Musk called Stargate vaporware, and Microsoft CEO Satya Nadella threw cold water on the hype, questioning whether OpenAI has the capital to pull this off. While Altman shot back with a simple “wrong,” the bigger question remains: Is this power consolidation accelerating innovation—or locking others out?
🔥 Hot Take:
Stargate isn’t just about compute—it’s a strategic moat for OpenAI and its partners. But does this scale open up AI access, or just concentrate control further? DeepSeek just proved that foundation models can be built efficiently outside of Big Tech. The real test: Can regional AI players in Europe, Latin America, and beyond follow suit?
⚡ DeepSeek’s Playbook: The Open-Source Shakeup
DeepSeek’s open-source AI model didn’t just challenge the status quo—it reset the playing field.
🔹 150 engineers. A fraction of the cost. And yet, DeepSeek-R1 is competing with proprietary models that cost billions.
🔹 Jevons Paradox in action – The more accessible AI gets, the more demand for compute actually increases.
🔹 The geopolitics of AI are shifting – The U.S. has Stargate, but DeepSeek signals China’s intent to lead in open AI development.
💡 Why This Actually Matters:
• AI infrastructure is decentralizing – DeepSeek proved you don’t need a trillion-dollar war chest to build powerful LLMs.
• Compute isn’t dead, it’s just different – Efficiency doesn’t reduce demand; it accelerates it.
• Regulatory roadblocks ahead? – With the EU AI Act coming online, transparency around training data will be a key battlefront.
• Security isn’t an afterthought – DeepSeek’s recent data exposure raises questions about trust at scale. Open-source AI lowers barriers, but can enterprises rely on it without stronger safeguards?
🔥 Hot Take: DeepSeek disrupted the assumption that AI dominance = massive spend. If OpenAI and NVIDIA are the incumbents, DeepSeek just opened the door for new challengers—not just in China, but anywhere with the right talent and compute access.
The real test? Whether open, AI-driven innovation can match Big Tech on security and trust—because access alone isn’t enough.
📖 Read More: My DeepSeek analysis from Monday
💰 The Money Moves: AI Investment Gets Focused
The era of “AI for AI’s sake” is fading fast. Investors aren’t just looking for compelling models—they want tangible, scalable results. The shift is clear: from general AI to vertical AI solving specific problems.
Biggest Bets This Week:
✅ ElevenLabs – New round at a $3B valuation (AI-powered voice synthesis)
✅ Skild AI: $500M in talks (robotics foundation models)
✅ Alice & Bob – $104M for quantum computing, aiming for commercial cat qubits by 2030 (long-term bet on AI’s future compute needs)
Early-Stage Startups to Watch:
🚀 Origami Agents: $2M seed (AI-powered sales teams)
🔍 Delve: $3M (AI for compliance evidence tracking)
🛡️ Conifers.ai: $25M (AI-driven cybersecurity)
🌾 Bonsai Robotics: $15M (AI-powered farming automation)
💬 Palona AI: $10M seed (AI for customer engagement)
🔥 Hot Take:
A couple of weeks ago, AI video startups were getting all the love. Now? The money’s shifting—healthcare, security, compliance, and robotics are pulling ahead. The hype hasn’t vanished, but investors are thinking bigger. AI that makes industries actually work is where the action is. If you’re building, that’s the signal to watch.
⚡ NVIDIA’s COSMOS: A Defensive Move or Next Big Leap?
DeepSeek’s open-source model just shook up the AI race, proving that you don’t need billions to build powerful LLMs. But where does that leave NVIDIA? Enter COSMOS—their next-gen AI compute platform, unveiled at CES earlier this month.
💡 Why This Still Matters:
• NVIDIA isn’t just a chipmaker anymore—COSMOS signals a bigger play to own AI infrastructure end to end.
• The Efficiency Paradox – As AI models get cheaper to train, demand for compute isn’t slowing—it’s accelerating.
• The Real Question – With DeepSeek proving open-source models can be built more efficiently, does COSMOS reinforce NVIDIA’s dominance or reveal its vulnerabilities?
🔥 Hot Take:
DeepSeek reshaped the narrative on who gets to lead in AI, proving that massive spend isn’t the only path to power. COSMOS is NVIDIA’s answer—ensuring that even as training gets more efficient, AI still runs through their ecosystem. Given how much AI innovation still relies on compute, that’s a smart bet.
But as AI infrastructure decentralizes and regional players step up (U.S., China, Europe, Middle East, Africa, India, Asia Pacific), does this cement NVIDIA’s leadership—or signal the rise of a more distributed AI future?
🦾 AI’s Leap into the Physical World
For years, AI was about chatbots, images, and text. Now? It’s going physical. The robotics wave is here.
💡 What’s Happening?
• SoftBank eyes a $500M investment in Skild AI—foundation models for robotics
• Figure AI and Physical Intelligence arepushing humanoid robots into real-world use cases
🔥 Hot Take:
AI’s biggest impact won’t be on screens—it’ll be in the real world. The move from digital workflows to physical automation is where AI starts to look like magic. The question is, will AI-powered robotics scale, or are we repeating the over-promises of past automation waves?
🧬 AI in Healthcare: Real Impact, Real Funding
Image Credits: Retro Biosciences
The pattern is clear: healthcare AI isn’t just hype—it’s attracting serious capital.
✅ Quibim’s $50M for AI-driven medical imaging
✅ VideaHealth’s $40M for AI-powered dentistry
✅ SafelyYou’s $43M for AI-assisted elder care
✅ Retro Biosciences’ $1B bet on extending human lifespan
🔥 Hot Take:
AI in healthcare isn’t about cutting costs—it’s about expanding access and unlocking treatments we couldn’t reach before. The real battle? Getting AI into real-world clinical settings at scale.
⚖️ AI Regulation & Reality Check
🔹 DeepSeek’s Security Breach Raises Big Questions – A publicly exposed database leaked over a million lines of log streams, including chat history and sensitive internal data.
🔹 OpenAI launches ChatGPT Gov (just as Trump revokes Biden’s AI executive order)
🔹 UK regulators scrutinize AWS/Microsoft cloud dominance
🔹 ABC News writers secure AI job protections
🔹 US Navy warns on DeepSeek models
💡 Why This Matters:
Three weeks in, and we’re watching policy shift in real time. The rules of AI are being written on the fly, with governments, unions, and companies all pulling in different directions.
🔥 Hot Take:
The “move fast, break things” era of AI just hit a reality check. Governance, security, and compliance aren’t just buzzwords—they’re becoming the deciding factors in who leads and who gets left behind.
DeepSeek’s security lapse is a warning shot: Open models might be the future, but trust and control still matter. The next phase of AI won’t just be shaped by who builds the best models—but by who ensures they’re secure, compliant, and enterprise-ready.
📝 Final Thoughts: AI’s Inflection Point
We started this journey talking about AI as the “tip of the spear.” Three weeks in, the shape of that spear is coming into focus—but let’s be real, the target keeps moving.
Every week, a new announcement shakes up the landscape—Stargate, DeepSeek, COSMOS, multi-billion-dollar funding rounds—and we’re left asking:
Not sure if we’re at the AI inflection point… or just another week in the GenAI rollercoaster.
✅ Infrastructure Wars: Stargate, COSMOS, and the race to own AI’s foundation
✅ Vertical AI is winning: General-purpose AI is giving way to industry-specific solutions
✅ Physical AI is here: Robotics and automation are moving from R&D to real-world deployment
✅ Regulation is real: AI’s next battleground isn’t just technical—it’s legal, political, and moving faster than expected
👀 One thing’s certain—buckle up, because next week is guaranteed to bring another plot twist.
