News
3 min read
AMD reached a stock market valuation of $1 trillion on 21 September 2026, driven by sustained demand for AI chips and sharply higher quarterly results. The company thus became the fourth American chipmaker to cross this threshold, after Nvidia, Broadcom and Micron.
On the same day, Belgian cybersecurity company Aikido released a new product that runs AI models locally within customers' own infrastructure. The two announcements mark a broader shift in the market: on the one hand, the consolidation of AI hardware suppliers; on the other, a growing need for privacy-conscious AI applications in sectors such as financial services.
In the second quarter of fiscal year 2026, AMD posted revenue of $11.5 billion, an increase of 50 percent compared with a year earlier. Data Centre products, including AI accelerators, accounted for $6.7 billion of that total. Earnings per share rose 82 percent to $1.66.
AMD's share price has climbed more than 180 percent over the course of 2026. AMD also announced a 10 percent price increase for its products, indicating a strong market position relative to customers dependent on AMD hardware for AI workloads.
AMD closed a multi-billion-dollar deal with OpenAI in October 2025, covering several years of collaboration including the joint construction of 6 gigawatts of compute capacity. Around the same time, a $1 billion partnership with the US government was announced for the construction of two supercomputers based on AMD's AI chips. The first, named Lux, is expected within six months of the announcement; the second, Discovery, is scheduled for 2029. A partnership with AI company Anthropic was also recently announced.
At the Advancing AI 2026 event on 23 July, AMD announced its entry into what the company calls 'physical AI': robotics, industrial automation and embedded systems. To that end, AMD introduced the Ryzen AI Embedded X100 Series processors, a Kria AI system-on-module and the AMD Robotics Partner Network.
AMD estimates that the market for AI silicon targeting physical applications will exceed $200 billion by 2035, according to the company's own projections. Whether that estimate proves accurate will depend in part on the pace of adoption in industrial sectors, where investment cycles are typically longer than in the data centre market.
Belgian company Aikido launched Altar on 21 September 2026, an AI model for cybersecurity that runs entirely on-premises at the customer's site. The model is derived from GLM-5.3, an open-weight model from Z.AI, and is designed for analysing sensitive code without that code ever reaching an external AI service.
Technically, Aikido has heavily compressed the original model through quantisation and expert pruning. Of the 256 routed experts in GLM-5.3, Altar retains 168. The stored weights shrank from approximately 1,507 gigabytes to 328 gigabytes in the process. Altar can be deployed on a node with four H200 GPUs, which lowers the implementation threshold compared with the full base model.
Aikido intends to integrate Altar into its own products. Aikido's customers include Belgian bank Belfius. The company reached a valuation of $1 billion in January 2026, following a $60 million Series B funding round. On 17 September 2026, Aikido received the AWS Security Competency certification for application security.
Aikido's move connects with a broader debate in Europe around data sovereignty in AI applications. Financial institutions and governments are under increasing pressure to account for which data they process through external AI services. A locally running model that is manageable in terms of weight offers a practical answer to that question, although the quality trade-off relative to larger cloud models remains a consideration for buyers.
For European founders and investors in the AI security market, Aikido's approach demonstrates that there is commercial space in compressing and adapting open-weight models for specific, regulated sectors. The combination of unicorn status, a concrete customer in Belfius and a technically substantiated product launch gives the company a solid starting position in that segment.
This article draws in part on the following sources.
Innovation Industries currently manages approximately 900 million euros across three funds and focuses on deeptech companies in the Benelux and Germany. The fund was founded in 2017 by four partners with backgrounds in engineering, law and operations.
On 29 September 2026, OpenAI launched two new products at DevDay in San Francisco: Dots, persistent AI agents that keep working after the user has logged out, and GPT-6.1 Sol, a lower-cost variant of GPT-6 with near-Astra performance for coding and professional work.
Enatom, a Groningen spin-out of the UMCG that makes photorealistic 3D anatomy models for medical education, raised €900,000 in April 2024. LUMO Labs, NOM and UNETI Ventures are jointly investing in the further development of the platform and the team.
8:08
34:00
21:31
19:41
16:27
21:52
1:14
0:52
