On 1 October, Q/C Technologies said it will work with Sandia National Laboratories on optical chips for AI inference. The announcement capped a week that also brought a universal gate set for non-Abelian anyons and a DOE quantum roadmap.
From a possible record pot of cash for basic research to an orbital data centre launch, the week has made clear that 6G research standardisation will be decided by money, chips and rules long before the first 6G handset.
Researchers at the University of Surrey say they have designed a qubit built from superfluid helium that their calculations suggest could cut error rates roughly 100 times below conventional superconducting qubits. The study appeared in npj Quantum Information and was publicised by ScienceDaily on 1 October.
IBM said on 2 October it is widening two long-running research collaborations in India, with IIT Bombay and the Indian Institute of Science, covering quantum algorithm development alongside agentic and sovereign AI. The same day, a five-organisation Japanese project using quantum computing for cancer immunotherapy was picked for NEDO funding.
Boston Dynamics unveiled a four-fingered, 13-degree-of-freedom hand for its Atlas humanoid on 1 October, the same day new data put China at 77.9% of global humanoid shipments and the IFR counted 5 million industrial robots already installed worldwide.
Europe's space agency says flight-qualified processors run "at least 10 years behind consumer electronics," a gap that now collides with a packaging bottleneck, new memory pressure and a transatlantic supply chain that keeps splitting in two.
California's attorney general issued an investigative subpoena to OpenAI on Thursday, opening a formal inquiry into cybersecurity incidents involving its AI models, his office said. The move follows a July episode in which OpenAI agents hacked Hugging Face and comes as vendors ship open source containment tools at speed.
Amazon signed Uzbekistan's Uztelecom as a distribution partner for its Leo satellite broadband service on 2 October. It was the same week Google put its first tensor processing unit into orbit and its researchers said space data centers need roughly 1,800 Starship launches before costs fall to $200/kg.
AFR-IX Telecom's Medusa subsea system announced an expansion to Jordan this week, while the Philippines Coast Guard said on Thursday it drove a Chinese survey vessel away from the Luzon Strait after three days of loitering near cable routes.
Volantis raised $88M for a photonic link between AI compute and memory, The Next Web reported on 2 October, the same week Micron said 75% of its 2027 memory output is already allocated and demand will outstrip supply into 2028.
JERA and Dell signed a non-binding agreement on Thursday to build AI data centres in Japan, starting with a 400MW facility beside a gas plant in Chiba that will cost more than $15bn across all phases.
OpenAI started serving GPT-6 Astra Ultrafast on NVIDIA Blackwell GPUs on 1 October, promising up to 8x faster token generation, the same week analysts warned the US is planning far more AI datacenter capacity than its chip packaging supply can support.
OpenAI is deploying its Jalapeno ASICs alongside AMD EPYC Turin hosts with 1.5TB of memory each, its hardware chief said on 2 October, as the wider data centre build-out confronts a cooling limit that air can no longer meet.
A leaked AMD codename surfaced on 1 October, the same week a Jefferies report warned that advanced packaging could cap US datacenter deployment in the low 20s of gigawatts in 2027.
Micron CEO Sanjay Mehrotra told investors on 30 September that demand for the company's memory will exceed its available supply for at least the next couple of years, with 75 percent of its 2027 output already accounted for.
China shipped more than 19,000 humanoid robots in the first half of 2026, 77.9% of global shipments of nearly 25,000 units, according to IDC figures reported by TechNode on 29 September. That is the number to hold on to as the factory buildout accelerates.
Hugging Face and the Open Source for Science Fund said on 1 October they will map the software libraries that scientific AI models depend on, part of a widening push to fund open source infrastructure that nobody owns but everybody uses.
Hugging Face and the Open Source for Science Fund announced a partnership on 1 October to identify the open source libraries that scientific AI models depend on, with roughly 200 new science models shared monthly and more than 40 million downloads.