How artificial intelligence is reshaping markets, companies, and the global economy.
Two startups — Lumotive and Neurophos — are applying optical metamaterial science to AI data center infrastructure. Lumotive debuted a programmable metasurface microchip on 19 March that can replace multiple optical switching components with no moving parts, targeting commercial launch by end of 2026. Neurophos claims its metamaterial optical modulators will deliver 50 times the compute density and energy efficiency of Nvidia's Blackwell GPU, with hyperscaler evaluations planned for this year and first systems targeted for early 2028.


Tower Semiconductor and Scintil Photonics have announced the world's first single-chip dense wavelength division multiplexing (DWDM) light engine designed for AI infrastructure. The LEAF Light photonic integrated circuit integrates lasers directly onto a co-packaged optics chip, enabling multiple wavelengths per fiber and up to 1.6 terabits per second over a single fiber — a development that could significantly cut latency and power consumption in GPU-dense data centers. Mass production is planned by end of 2026.
German-Chinese robotics manufacturer **Kuka AG** has warned that Europe's industrial companies are adopting artificial intelligence too slowly, leaving them vulnerable to faster-moving rivals in the US and Asia. The company is now prioritising growth in those markets as European factory modernisation stalls. The warning from one of the world's leading industrial robot makers adds senior industry weight to growing concerns about Europe's competitiveness in advanced manufacturing.
Super Micro Computer has convened a committee of independent directors to investigate how two employees and a contractor came to be indicted over allegedly selling servers to China. The committee has retained a law firm to lead the probe. The case raises fresh scrutiny over export controls and compliance practices at one of the world's largest server manufacturers, a critical supplier to AI data centre operators.
Broadcom shares climbed after the company confirmed a chip development partnership with Google centred on Tensor Processing Units (TPUs), positioning Broadcom as a credible alternative to Nvidia in the AI accelerator market. The deal signals Google's continued push to diversify its AI chip supply chain beyond dominant incumbent suppliers.

A new MIT Technology Review analysis argues that companies must redesign their core workflows around AI agents rather than layering them onto existing systems. Unlike rules-based automation, AI agents can learn, adapt, and execute entire workflows autonomously — but only when processes are built with agents at the centre. The piece warns that bolting agents onto fragmented legacy infrastructure using traditional optimisation methods will leave most of their potential unrealised.
Microsoft is struggling to keep pace with rivals in the global data center build-out after a period of reduced capital expenditure, according to Bloomberg Technology. The slowdown has left the company scrambling to catch up as demand for AI infrastructure accelerates, raising questions about its ability to compete with hyperscalers that maintained aggressive expansion plans.

Amazon Web Services and maritime intelligence firm Windward have published details of a generative AI system that transforms how analysts investigate vessel anomalies at sea. Rather than responding to isolated alerts, analysts now receive contextual, AI-assembled intelligence packages — reducing the time spent on data collection and accelerating decision-making. The system, detailed in an AWS Machine Learning blog post, combines geospatial data with agentic AI built on AWS infrastructure.

Cisco CEO Chuck Robbins has stated publicly that orbital data centers are not just plausible but inevitable, aligning himself with Elon Musk's ambitions over Sam Altman's scepticism. Speaking on The Verge's Decoder podcast, Robbins said Cisco teams began assessing space-ready networking infrastructure roughly two to three months ago. He also acknowledged that the current AI infrastructure boom has bubble-like characteristics — a notable admission from the CEO of a company that was briefly the world's most valuable during the dot-com era.

Artificial intelligence is changing how small e-commerce sellers identify, develop, and launch products, according to a report by MIT Technology Review. Tools like Alibaba's Accio are helping independent sellers analyse market demand and supplier options at a scale previously available only to large retailers. The shift signals a broader democratisation of data-driven decision-making in e-commerce, with significant implications for how small businesses compete.

Advanced chip packaging — the engineering discipline that connects multiple chips into a single, high-performance unit — has emerged as a critical battleground in the AI hardware race. Intel is making it a cornerstone of its strategy, betting that its packaging technology can attract AI customers and generate billions in revenue as demand for faster, denser compute continues to accelerate.
Hon Hai Precision Industry, the Taiwanese manufacturer and key Nvidia supply chain partner, reported a 29.7% rise in quarterly sales, meeting analyst estimates. The results signal that enterprise demand for AI infrastructure remains robust even as geopolitical tensions — including conflict in the Middle East — create broader economic headwinds.

Anthropic is cutting off third-party AI harness OpenClaw from Claude subscription limits starting April 4th. Users who want to continue using OpenClaw with Claude must switch to a separate pay-as-you-go billing model, a move that effectively prices out casual users. The policy shift comes as OpenClaw's creator, Peter Steinberger, now works for rival OpenAI.
Stay informed
Get DeepBrief delivered to your inbox.