• Playlist
  • Seattle Startup Toolkit
  • Portfolio
  • About
  • Job Board
  • Blog
  • Token Talk
  • News
Menu

Ascend.vc

  • Playlist
  • Seattle Startup Toolkit
  • Portfolio
  • About
  • Job Board
  • Blog
  • Token Talk
  • News

Token Talk 27: Sovereignty as a Service

July 24, 2025

Jensen Huang made his pitch.

The CEO of the world’s first $4 trillion company met with President Trump and pushed to restart Nvidia chip sales to China, challenging the U.S.’s hawkish AI stance. He won over Commerce Secretary Howard Lutnick, who said the goal is to get Chinese developers “addicted to the American technology stack.”

Shortly after, Nvidia announced it would resume H20 chip sales to China while ramping up U.S. production with a $500 billion investment. 

OpenAI’s massive datacenter deal in the UAE and Nvidia’s freshly unbanned H20s reflect Washington’s new playbook: sell the shovels, keep the mines — awesome for Nvidia, awful for AWS, catnip for every sovereign startup. 

From Sweden to Singapore, governments are racing to bottle lightning before the price of a prompt hits zero. Beneath the concrete and coolant is a simple truth: AI independence is national sovereignty. Forget gold reserves and space programs. Power today is measured literally in gigawatts and petaflops per capita. The irony is that true digital sovereignty now depends on one American company. For this plan to work, Nvidia must become the global supplier of independence.

And it's not just the UAE making large investments: 

  • Canada dropped $2B to make “strategic investments in public and commercial infrastructure.” 

  • Japan bought thousands of Nvidia H200s for its new AI Bridging Cloud Infrastructure (ABCI 3.0) supercomputer. 

  • The French government is constructing a “cloud de confiance” (cloud of confidentiality) that is already hosting a version of Microsoft Azure. 

  • India’s IndiaAI Mission aims to democratize compute with 10,000+ GPUs and national datasets. 

  • xAI is partnering with Saudi Arabia to build and lease 200MW of data-center. 

  • And South Korea's 3GW data center, the world’s largest, is under construction. 

Stateside, the AI buildout is in full swing. In Pennsylvania alone, Trump recently touted more than $90 billion in new AI and energy investments. Hyperscalers like Google, CoreWeave, and Meta are in a frantic race to secure grid capacity. 

It’s easier to conceptualize these companies’ growth by their energy demands. Digital factories already consume more than 4% of the nation's electricity, a figure projected to triple to 12% by 2028.That is playing out in real-time in northern Virginia, where the construction of 30 new data centers in just two years has “Data Center Alley” residents facing a 50% increase in electricity bills. 

But the promise of “AI sovereignty” keeps countries pushing ahead. 

Here’s my take on the news: 

The age of relying on someone else’s cloud (at least for the government) is over.

The same hyperscalers begging for megawatts could quietly become the Pentagon’s favorite Trojan horse. Every one of those gleaming new data centers comes pre-wired with a backdoor handshake to Washington. No, not in the crude sense of a literal switch, port, or door, but in the invisible way that modern sovereignty works: through standards, firmware, and the quiet insertion of compute modules that answer to a .gov root certificate before they answer to anyone else.

The U.S. doesn’t need to own every rack or risk; it just needs to own the trust anchor. A single line of code in the boot ROM — signed by NIST, blessed by CISA, and baked into every H200, B200, and whatever comes next. The compute is sovereign, the power is local, but the keys are quietly federated back to the U.S.

The house always wins, or something like that. 

Call it “cloud capitalism with American characteristics.” Washington gets to project soft power and secure its AI supply chain without a single line item in the defense budget. Abu Dhabi gets the keys to a 21st-century economy and a permanent alliance with U.S. corporate interests. And the American taxpayer? They get to watch their 401ks swell as Big Tech books tens of billions in revenue, all while footing precisely zero of the bill for this new global security architecture.

Of course, the plan only works as long as nobody calls the bluff. The moment a major power or a coalition of non-aligned nations develops a viable, open-source alternative — a 'de-Americanized' chip or stack — this entire strategy collapses. Washington lifted the Nvidia ban, but Beijing isn’t taking any more chances. Xi’s response was the Huawei Ascend 910 AI chip. 

Still Nvidia shares bounced 9% on the news. 

If the U.S. leverages the standards and firmware in its exports, it could accelerate the very 'splinternet' we fear, turning its current market dominance into a long-term strategic liability as the world rushes to build a truly sovereign alternative.

Alternatives that will most certainly be new startups. Like the Slovakia-based Tatra Supercompute that gives its EU customers access to powerful clusters of NVIDIA H100s — all GDPR compliant. Or Lumina CloudInfra, which is focused on building India's sovereign AI platform, defining the standards and digital public infrastructure needed to ensure the nation's data and AI destiny are its own. And Seattle-based Hedgehog (Ascend portco) is connecting everything together with its open-source, Kubernetes-native fabric that automates the deployment of high-performance, inference networks. Essential for anyone building their own cloud.

As someone who gets a kick out of running models directly in the browser using WebGPU, or on my laptop, I've seen a glimpse of a decentralized future where the compute happens at the edge, on the user's terms. So I understand these countries' desire for local digital control. It’s the same impulse that drives a developer to spin up a home lab instead of swiping a corporate credit card on AWS. 

The grand American strategy is a bet that convenience will always trump sovereignty. A bet that "it just works" is more powerful than "we own it." But that bet only pays off until the moment it doesn't. So stay tuned!

Tags Token Talk, Nvidia, China AI

Image generated in OpenAI’s new image generation feature, with the prompt: “Create a headline image in Studio Ghibli style of this article.”

Token Talk 10: What Startups Gain from China’s AI Push

March 26, 2025

By: Thomas Stahura

The race to dominate artificial intelligence is accelerating on every front, as research labs across the globe push full throttle on new model releases while governments move to cement AI supremacy. 

In the past few weeks, Google released two major models, OpenAI launched long-awaited image capabilities, and Chinese labs pushed open-source systems that rival the best from the West. What began as a battle between private research labs is now a global competition shaped by open models, national strategies, and shifting power dynamics. 

Here's a breakdown of what just happened:

Google announced Gemma 3, the latest model in its Gemma trilogy. At around 27 billion parameters, I wouldn’t call it “small,” yet it punches above its weight class. It’s the only open model that can take video as input. Mistral open-sourced Mistral-Small-3.1 a few days later, a 24 billion parameter model that outperforms Gemma 3 on most benchmarks.

But really the larger news here is Gemini 2.0 Flash Experimental. Google’s new closed-source flagship model and the company's first unified multimodal model. Meaning, the model can generate and understand both images and text simultaneously. I’ve been playing around with it. It is capable of editing images using simple text prompts, generating each frame of a GIF, and even composing a story complete with illustrations. (This is similar to Seattle startup 7Dof, which showcased a visual chain-of-thought editing tool at South Park Commons last year.) 

Traditionally, transformer models were used to generate text, while diffusion models generate images. Today, researchers are experimenting with unifying both architectures into a single model (similar to what is going on with VLA models in robotics). The ultimate goal is to build a model that unifies the text, image, and audio spaces.

Gpt-4o has had image generating abilities for a while. Greg Brokman demoed gpt-4o generating images in May. And this week the company finally launched the capability. 

At this point in the AI race, OpenAI seems to be reacting more than leading. Launching 4o’s image gen was a response to Gemini 2.0 Flash Experimental. 

Trump said multiple times he wants “American AI Dominance.” And, to that effect, the White House invited public comment on its AI Action Plan. OpenAI published its response, slamming DeepSeek and urging the administration to implement the following: 

  1. An export control strategy that exports democratic AI

  2. A copyright strategy that promotes the freedom to learn

  3. A strategy to seize the infrastructure opportunity to drive growth

  4. And an ambitious government adoption strategy.

Google also responded, urging America to:

  1. Invest in AI

  2. Accelerate and modernize government AI adoption

  3. Promote pro-innovation approaches internationally

China has their own plan. 

Dubbed the “New Generation Artificial Intelligence Development Plan” (2017), the agenda aims to make China the global leader in AI by 2030. The worry seems to be about the sheer quality and openness of the models out of China today. It’s hard to name a model out of a Chinese AI lab that isn’t open source. 

Over the course of a week earlier this month, DeepSeek open-sourced all technical details used in the creation of its R1 and V3 models. All except for the actual dataset used to train the models (adding to the suspicion that DeepSeek trained on gpt-4o outputs). 

DeepSeek also open-sourced Janus-Pro. Though the model got significantly less attention than its big brother, Janus-Pro is a unified multimodal model (like Gemini 2.0 Experimental), capable of generating and understanding both images and text — one of the first open-source models of its kind.

Qwen, the AI lab out of Alibaba Cloud, has launched its own reasoning model: QwQ-32B, competing with and reaching DeepSeek R1 performance on many benchmarks. The model already has 615k downloads on Hugging Face.

OpenBMB (Open Lab for Big Model Base) is a Chinese AI research group out of Tsinghua University. The group is most known for MiniCPM-o-2_6, a unified multimodal model capable of understanding images, text, and speech, as well as generating text and speech. The model is at gpt-4o levels, according to the benchmarks, and has 766k downloads.

DeepSeek V3.1 also launched this week. The model leapfrogged Grok 3 and Claude 3.7 to become the best performing non-resoning model. The first time an open-source model achieved SOTA. 

That is until Google 2.5 Experimental dropped a few hours later. More on that next week. 

Ok, here’s my take on the flood of releases: 

This is good news for startups, full stop. More models means more competition, and that means lower prices. Even if the U.S. bans Chinese models, most are fully open. Developers can fine-tune them and build whatever they need.

The real challenge now is the viability of America’s top AI labs. If China can flood the market with cheap, open, high-quality models, they could undercut their U.S. counterparts. It’s a familiar playbook — one China used before in other industries. This time, it’s electrons instead of atoms. That shift might tilt the board in China’s favor.

Only time will tell, so stay tuned!

Tags Token Talk, China AI, OpenAI, AI

FEATURED

Featured
Metro Multiples: A ranking of top startup ecosystems by return on investment
Metro Multiples: A ranking of top startup ecosystems by return on investment
Mapping Cascadian Dynamism
Mapping Cascadian Dynamism
Subscribe to Token Talk
Subscribe to Token Talk
You Let AI Help Build Your Product. Can You Still Own It?
You Let AI Help Build Your Product. Can You Still Own It?
Startup-Comp (1).jpg
Early-Stage Hiring, Decoded: What 60 Seattle Startups Told Us
Booming: An Inside Look at Seattle's AI Startup Scene
Booming: An Inside Look at Seattle's AI Startup Scene
SEATTLE AI MARKET MAP V2 - EDITED (V4).jpg
Mapping Seattle's Enterprise AI Startups
Our 2025 Predictions: AI, space policy, and hoverboards
Our 2025 Predictions: AI, space policy, and hoverboards
Mapping Seattle's Active Venture Firms
Mapping Seattle's Active Venture Firms
PHOTOS: Founders Bash 2024
PHOTOS: Founders Bash 2024
VC for the rest of us: A big tech employee’s guide to becoming startup advisors
VC for the rest of us: A big tech employee’s guide to becoming startup advisors
Valley VCs.jpg
Event Recap: Valley VCs Love Seattle Startups
VC for the rest of us: The ultimate guide to investing in venture capital funds for tech employees
VC for the rest of us: The ultimate guide to investing in venture capital funds for tech employees
Seattle VC Firms Led Just 11% of Early-Stage Funding Rounds in 2023
Seattle VC Firms Led Just 11% of Early-Stage Funding Rounds in 2023
Seattle AI Market Map (1).jpg
Mapping the Emerald City’s Growing AI Dominance
SaaS 3.0: Why the Software Business Model Will Continue to Thrive in the Age of In-House AI Development
SaaS 3.0: Why the Software Business Model Will Continue to Thrive in the Age of In-House AI Development
3b47f6bc-a54c-4cf3-889d-4a5faa9583e9.png
Best Practices for Requesting Warm Intros From Your Investors
 

Powered by Squarespace