Ep 823- The U.S. Vs China AI Cold War Is Starting- What It Means and How It Impacts You

AI Leadership Is Geopolitics Not Just Tech
- The U.S. vs China AI conflict is becoming a geopolitical cold war centered on economic growth, standards, and leverage.
- Jordan Wilson argues AI leadership yields economic compounding, military leverage, and global rule-setting power. (Time 0:00:16)
Chinese Models Closing Frontier Parity
- Chinese open models have rapidly closed the capability gap and are now in the same upper echelon as Western frontier models.
- Recent releases like Kimmi K3 and Qwen 3.8 show near-parity and faster public rollout than expected. (Time 0:03:41)
Open Weights As Strategic Infrastructure
- China intentionally released open weights to become the default base other countries build on and to capture downstream revenues.
- Jordan highlights narrative control as another motive: models avoid sensitive topics and shape future training data. (Time 0:09:01)
How GLM 5.2 Sparked Rapid Catchup
- Jordan recounts the timeline: GLM 5.2 started the shift, then Kimi K3 and Qwen 3.8 accelerated parity to within weeks.
- He notes Chinese models moved from six-month lag to one-two months behind US leaders. (Time 0:12:51)
Trillion Parameter Models Break Consumer Runability
- Multiple-trillion-parameter open models mean you can no longer run frontier weights on consumer hardware.
- Jordan explains enterprises now need racks of GPUs or small data centers to run these open models locally. (Time 0:17:11)
Price Per Task Favors Closed Western Models
- Cost per task now favors closed US models: OpenAI, Grok, and Meta sit in the sweet spot of high intelligence and lower cost.
- Jordan cites Artificial Analysis benchmarks where Chinese open models lost the cost-effectiveness quadrant. (Time 0:20:57)
Distillation Speeds Progress But Triggers Pushback
- Distillation (copying question-answer behavior from frontier models) accelerated Chinese progress but doesn’t fully explain improvements.
- Jordan notes distillation is legally and politically fraught and is prompting export-control responses. (Time 0:25:21)
Microsoft Tested Chinese Models For Cost Routing
- Jordan describes Microsoft’s reported evaluation of DeepSeek as a cost-saving backup for Copilot routing.
- He suggests enterprises will use model routing by cost, capability, security, geography, and regulation. (Time 0:28:18)
Inventory Chinese Models And Prepare Swap Plans
- Do inventory and document every Chinese open model in use and route tasks by cost, security, and policy risk.
- Jordan advises preparing swap-out plans and being ready to shift workloads if export controls or bans harden. (Time 0:31:15)