ByteDance Negotiates Major AI Chip Deal with China's Iluvatar CoreX
1h ago|4 min readStandard
Fazen Markets Editorial Desk
Collective editorial team · methodology
artificial-intelligencesemiconductors
Fazen Markets Editorial Desk
Collective editorial team · methodology
Trades XAUUSD 24/5 on autopilot. Verified Myfxbook performance. Free forever.
Risk warning: CFDs are complex instruments and come with a high risk of losing money rapidly due to leverage. The majority of retail investor accounts lose money when trading CFDs. Vortex HFT is informational software — not investment advice. Past performance does not guarantee future results.
ByteDance, the parent company of TikTok, is actively negotiating a significant purchase of artificial intelligence processors from domestic Chinese chipmaker Iluvatar CoreX. Multiple sources confirmed the advanced talks on 15 June 2026. This potential deal represents a major strategic shift for the social media giant as it seeks to secure a stable supply of high-performance computing hardware for its global AI ambitions, circumventing ongoing US export restrictions on leading Western alternatives.
The negotiation occurs amid intensified US scrutiny of China's technological advancement. In October 2022, the US Bureau of Industry and Security imposed sweeping export controls, severely limiting Chinese companies' access to advanced AI chips from NVIDIA and AMD. The current macro backdrop features heightened geopolitical tensions and a global race for AI supremacy. China's government has responded with substantial subsidies and policy support for its domestic semiconductor industry, aiming to achieve technological self-reliance. ByteDance's move is a direct adaptation to this new reality, prioritizing supply chain security over peak performance.
This pivot to a domestic supplier follows a pattern set by other Chinese tech giants. In late 2025, Alibaba Group committed over $1 billion to purchasing AI accelerators from Huawei's Ascend unit. Tencent Holdings similarly increased its procurement of chips from startup Enflame Technology by 300% year-over-year in Q1 2026. The triggering catalyst for ByteDance is the imminent launch of its next-generation large language model, which requires immense computational power for both training and inference, creating an urgent need for guaranteed hardware access.
Iluvatar CoreX's flagship Tiangai 270 accelerator is the likely subject of the negotiation. Benchmark tests show the Tiangai 270 delivers approximately 125 teraflops of FP16 performance, a key metric for AI training. This performance lags behind NVIDIA's restricted H100, which delivers 1,979 teraflops of FP16 performance with sparsity. The Chinese chip also consumes 225 watts of power, compared to the H100's 700 watts. A potential deal could be valued in the hundreds of millions of dollars, based on ByteDance's vast AI infrastructure needs.
| Metric | Iluvatar CoreX Tiangai 270 | NVIDIA H100 (for comparison) |
|---|---|---|
| FP16 Performance | ~125 TFLOPS | 1,979 TFLOPS |
| Power Consumption | 225W | 700W |
| Memory Bandwidth | 819 GB/s | 3.35 TB/s |
China's domestic AI chip market is projected to reach $7 billion in 2026, up from $5 billion in 2025, according to Sino Tech Insights. This represents a 40% growth rate, far exceeding the global average of 15%. Despite this growth, domestic suppliers collectively hold less than 15% of the Chinese market share for high-end AI training chips, which remains dominated by NVIDIA's pre-restriction inventory.
The most direct beneficiary is Iluvatar CoreX itself, which would gain a marquee customer and validation for its technology, potentially attracting further investment. Other Chinese AI chip designers like Cambricon Technologies (688256:SHH), Enflame, and Horizon Robotics also stand to gain from increased investor interest and a broader industry shift toward domestic sourcing. Semiconductor manufacturing equipment suppliers to these firms, such as NAURA Technology Group (002371:SHE), could see increased orders.
The primary risk is performance disparity. A cluster of Tiangai 270 chips would require significantly more units and consume more power to achieve computational results comparable to a smaller cluster of latest-generation Western GPUs, increasing ByteDance's operational costs and potentially delaying product development cycles. Institutional investors are increasing long positions in the CSI Semiconductor Index, which is up 18% year-to-date versus the Shanghai Composite's 3% gain. Short interest remains elevated in US-listed Chinese tech ADRs like BABA and PDD, reflecting lingering concerns over geopolitical headwinds.
The next major catalyst is China's planned release of its "National Integrated Circuit Industry Investment Fund Phase III" in Q3 2026. The size and focus of this state-backed fund will signal the government's continued commitment to subsidizing domestic chip production. The US Department of Commerce's quarterly review of export control policies, due in late July 2026, is another critical event that could further tighten or loosen restrictions.
Key levels to monitor include the share price of Iluvatar CoreX's major backers, such as the Summitview Capital portfolio. The yield on China's National Integrated Circuit Industry Investment Fund bonds, currently at 3.2%, will reflect market confidence in the sector's state-backed financing. Should ByteDance finalize a deal, its capital expenditure guidance in its next earnings report on 20 August 2026 will provide concrete data on the financial scale of its domestic procurement shift.
Major players include Huawei's Ascend unit, which produces the Ascend 910 chip, and startups like Iluvatar CoreX, Cambricon, and Enflame. These firms have received billions in state and venture funding since 2020. Cambricon is publicly traded on Shanghai's STAR Market, while others remain privately held. Their technological progress is measured by their ability to close the performance gap with NVIDIA in terms of floating-point operations per second and memory bandwidth for large-scale model training.
The US export controls implemented in 2022 set two key thresholds: a maximum chip-to-chip I/O bandwidth of 600 GB/s and a maximum combined processing performance of 4,800 TOPS. Any chip exceeding either limit requires a special license for export to China, which is subject to a presumption of denial. This effectively blocks the sale of the most advanced data center GPUs designed for AI training, forcing Chinese firms to use older generations or develop domestic alternatives.
The gap remains significant but is narrowing. For example, NVIDIA's H100 GPU outperforms most Chinese alternatives in raw compute by a factor of 10-15x for certain AI workloads. However, Chinese chips are often optimized for specific AI models and can offer better performance-per-watt or cost efficiency for inference tasks. The gap is smallest in inference chips and largest in high-performance training chips, which require the most advanced semiconductor manufacturing processes.
ByteDance's potential deal accelerates China's strategic decoupling from Western semiconductor technology.
Disclaimer: This article is for informational purposes only and does not constitute investment advice. CFD trading carries high risk of capital loss.
Vortex HFT is our free MT4/MT5 Expert Advisor. Verified Myfxbook performance. No subscription. No fees. Trades 24/5.
Position yourself for the macro moves discussed above
Start TradingSponsored
Open a demo account in 30 seconds. No deposit required.
CFDs are complex instruments and come with a high risk of losing money rapidly due to leverage. You should consider whether you understand how CFDs work and whether you can afford to take the high risk of losing your money.