AI International Competition

State of play.

• China is converting DeepSeek from independent startup to state-backed national champion. China's state-backed Integrated Circuit Industry Investment Fund is leading DeepSeek's first external funding round at a reported $45 billion valuation—a quadrupling in two weeks—signaling Beijing's direct financial consolidation of its leading AI firm .
• The DeepSeek-Huawei axis is now a functional alternative supply chain. DeepSeek V4 launched on Huawei Ascend chips with API pricing at roughly one-ninth of OpenAI's rate, demonstrating that U.S. semiconductor export controls have not arrested Chinese frontier model development .
• The U.S. has escalated from export controls to diplomatic and enforcement posture on model distillation. The State Department issued a global diplomatic cable accusing DeepSeek, Moonshot AI, and MiniMax of IP extraction from U.S. models; the White House science adviser issued a parallel memo framing the activity as an "industrial-scale campaign" .
• The EU's high-risk AI compliance deadline is live and unresolved. Omnibus trilogue talks failed in April; if May negotiations do not close, the original August 2, 2026 deadline for high-risk AI system obligations takes effect unchanged .
• For counsel advising technology companies, AI developers, or clients with China exposure, the practical baseline is a multi-vector compliance environment: U.S. export controls on chips, a nascent sanctions/IP-theft enforcement regime targeting model distillation, Chinese outbound investment controls blocking cross-border AI acquisitions, and an EU compliance cliff arriving in August 2026.

Where things stand.

• DeepSeek is transitioning to state-backed status with governance implications that remain undisclosed. The Big Fund's lead investment has not disclosed governance terms, making it unclear whether DeepSeek's open-source posture survives or whether the investment triggers CFIUS-adjacent scrutiny from U.S. regulators .
• Huawei's domestic chip ecosystem is scaling rapidly. Huawei projects AI chip revenue of approximately $12 billion in 2026—up from $7.5 billion in 2025 per Financial Times reporting—anchored by mass production of the Ascend 950PR and a planned 950DT launch; major Chinese enterprises are actively securing Huawei chips as Nvidia alternatives .
• U.S. chip equipment export controls are expanding to new Chinese manufacturers. Commerce ordered Lam Research, Applied Materials, and KLA to halt shipments to Hua Hong and its subsidiary Huali Microelectronics, targeting facilities assessed capable of 7-nanometer production .
• China is using outbound investment controls to block AI talent and algorithm transfers to U.S. acquirers. Beijing blocked Meta's $2 billion acquisition of Manus by imposing exit bans on the founders, asserting that algorithms developed domestically constitute a prohibited technology export regardless of corporate domicile .
• Chinese geospatial AI is now an active military threat vector. U.S. intelligence assessed that MizarVision, a partially state-owned Chinese geospatial AI company, published AI-tagged satellite imagery of U.S. Middle East bases that Iran's IRGC used for targeting .
• The open-source vs. proprietary policy debate is unresolved and consequential. A USCC report found Chinese open-source models account for 41 percent of HuggingFace downloads versus 36.5 percent for U.S. models; the policy debate over whether the U.S. should embrace open-source to compete is active and influencing defense procurement and export control discussions .
• China's domestic AI governance imposes structural constraints on its own firms. The CAC requires generative AI to adhere to "core values of socialism" with programmer personal liability for violations—creating a compliance gap between open-source training data and enforced censorship that affects any firm with Chinese AI operations or partnerships .
• Allied AI competition is consolidating around "sovereign AI" frameworks. The Cohere-Aleph Alpha merger, backed by Germany's Schwarz Group and announced with government ministers from both Canada and Germany present, reflects a Canada-Germany Sovereign Technology Alliance explicitly designed to build AI capacity outside U.S. and Chinese dominance .
• U.S. frontier AI deployment in government is outpacing formal risk governance. Anthropic's Mythos model continued operating across U.S. agencies including the NSA despite the Department of War flagging Anthropic as a supply chain risk; the UK AI Security Institute tested Mythos and restricted access to eight European cyber agencies .
• American humanoid robotics depends on Chinese component supply chains. U.S. humanoid robot platforms—including Tesla's Optimus—rely heavily on Chinese-sourced components and suppliers, creating a structural supply chain vulnerability that intersects with AI hardware competition .

Latest developments.

• China's Big Fund leads DeepSeek funding round at $45 billion valuation, converting the startup into a state-backed national champion; governance terms and Tencent/Alibaba participation remain undisclosed .
• DeepSeek V4 launched with Huawei Ascend chip support, 1.6 trillion-parameter Pro variant, and API pricing at $3.48 per million output tokens—roughly one-ninth of OpenAI's rate—triggering a Chinese enterprise scramble for Huawei chips .
• State Department issued a global diplomatic cable accusing DeepSeek, Moonshot AI, and MiniMax of "extraction and distillation" of U.S. AI models; a separate demarche was sent to Beijing .
• White House science adviser Michael Kratsios issued a memo framing Chinese model distillation as an "industrial-scale campaign"; House Foreign Affairs Committee advanced legislation to sanction foreign actors extracting capabilities from U.S. closed-source models .
• Commerce Department ordered Lam Research, Applied Materials, and KLA to halt shipments to Hua Hong and Huali Microelectronics over advanced chip production concerns .
• EU Omnibus trilogue failed after 12 hours of talks on April 28; May negotiations are the last window before the August 2, 2026 high-risk AI compliance deadline takes effect unchanged .
• Beijing blocked Meta's $2 billion Manus acquisition via founder exit bans, asserting domestic algorithm development constitutes a prohibited technology export .
• Cohere and Aleph Alpha announced a $20 billion merger under the Canada-Germany Sovereign Technology Alliance, with government ministers from both countries attending the Berlin announcement .
• U.S. intelligence assessed that MizarVision's AI-tagged satellite imagery was used by Iran's IRGC to target U.S. military bases across the Middle East .
• Huawei projects AI chip revenue of approximately $12 billion in 2026, per Financial Times reporting, anchored by Ascend 950PR mass production .
• Britain is actively courting Anthropic for London expansion following the Pentagon supply-chain-risk clash .

Active questions and open splits.

• Whether model distillation constitutes actionable IP theft or sanctions-eligible conduct. The State Department cable and Kratsios memo treat distillation as an "industrial-scale" violation, but the legal theory is unsettled—distillation is a standard industry practice used by U.S. and Chinese firms alike. What enforcement mechanism attaches, and under what statute, is unresolved .
• Whether DeepSeek's state-backed ownership triggers U.S. regulatory response. The Big Fund's investment raises questions about whether existing CFIUS, export control, or sanctions frameworks apply to a company that is open-source, foreign-domiciled, and now state-owned—and whether U.S. enterprises can continue accessing DeepSeek APIs .
• Whether the Huawei-DeepSeek partnership constitutes sanctions evasion. DeepSeek is training and deploying on Huawei Ascend chips while Huawei remains under U.S. sanctions; the partnership's compliance status under existing export control frameworks has not been adjudicated .
• How far China's outbound technology controls extend to overseas-incorporated AI companies. The Manus block asserts jurisdiction over algorithms developed in China regardless of corporate domicile; the scope of this theory—and whether it applies to other Chinese-founded, Singapore- or U.S.-incorporated AI companies—is undefined .
• Whether the EU high-risk AI compliance deadline will be extended or take effect in August 2026. The Omnibus stalemate leaves companies facing either an August cliff or a last-minute deal; the compliance posture for embedded high-risk systems (medical devices, regulated products) is particularly uncertain .
• Whether dual-use geospatial AI companies face new export control or liability exposure. The MizarVision-IRGC targeting chain raises the question of whether commercially available AI tools used for military targeting against U.S. forces trigger existing export control frameworks, new designations, or novel liability theories .
• Whether the open-source vs. proprietary policy split will produce conflicting regulatory signals. The USCC data showing Chinese open-source dominance on HuggingFace is driving a policy argument for U.S. open-source promotion—directly in tension with export control and IP-protection rationales for restricting model access .

What to watch.

• May EU Omnibus trilogue outcome—a failed deal means August 2026 high-risk AI compliance obligations take effect for all covered systems with no delay.
• Whether the Trump administration translates the Kratsios memo and State Department cable into concrete sanctions designations, export restrictions on model access, or terms-of-service enforcement actions against Chinese AI firms.
• Whether U.S. regulators respond to the Big Fund's DeepSeek investment with CFIUS scrutiny, entity list additions, or restrictions on U.S. enterprise access to DeepSeek APIs.
• Closing and regulatory clearance status of the Cohere-Aleph Alpha merger—the first major "sovereign AI" consolidation transaction, which will test how regulators treat geopolitical alignment as a merger rationale.
• Whether Commerce expands chip equipment export restrictions beyond Hua Hong to additional Chinese manufacturers approaching advanced-node production capability.
• Trump-Xi summit outcomes on semiconductors and IP—any bilateral agreement or breakdown will directly reshape the enforcement landscape for model distillation claims and chip export controls.