HKT plans a 3.2 Tbps data-centre route from Lok Ma Chau to Tseung Kwan O by the end of 2026, but has not disclosed customer bandwidth, price, project cost or end-to-end latency, while Sandy Ridge's developer expects to start within 42 months of its land award.
HKT has put a capacity figure, a route and a deadline on its next AI network upgrade. It has not yet provided enough commercial or technical detail to show that separate data centres will perform like one computing cluster.
HKT said in a company announcement distributed on July 17 that it plans to roll out what it calls Hong Kong's first 3.2 Tbps AI Data Centre Inter-connect Superhighway. The first phase is intended to connect Lok Ma Chau with Tseung Kwan O by the end of 2026.
That wording matters: this is a plan to launch, not a completed 3.2 Tbps service. HKT says the network will connect major data-centre clusters with its new AI Exchange in the Lok Ma Chau Loop and support Remote Direct Memory Access, RoCEv2, Multipath Reliable Connection and Ultra-Ethernet Transport. It says infrastructure providers, universities and research laboratories will be able to pool computing capacity for AI training, inference and scientific simulations.
Those workload benefits are HKT's claims. The release provides no route length, customer-facing port speed, definition of whether 3.2 Tbps is aggregate or directional capacity, service-level commitment, named customer or committed traffic. It also gives no price, construction cost or utilization forecast.
The staging has also been blurred in some coverage. One local headline suggested a Sandy Ridge connection by year-end, while its article body described the year-end phase as Lok Ma Chau to Tseung Kwan O and Sandy Ridge as a later extension. HKT's release says only that the group is prepared to extend coverage to Sandy Ridge.
The 3.2 Tbps number is four times HKT's earlier 800 Gbps headline, but the available disclosures do not establish a like-for-like service comparison.
When HKT introduced the 800 Gbps AI Superhighway in February 2025, it described 800 Gbps as the maximum bandwidth in each direction. It warned that experienced speed could be lower because of devices, technology, software, configuration and other factors. That service used optimized routes among Tseung Kwan O, Chai Wan and Kwai Chung, and HKT offered fixed-rental and usage-based payment options.
The new release does not repeat those definitions or disclose a tariff. A reader can compare the two headline numbers arithmetically, but cannot determine whether a customer can buy four times as much bandwidth on equivalent terms.
Nor does HKT enter an empty local market. HKBN announced in July 2025 that it was developing a 1.6 Tbps optical core and offering a 100 Gbps Metro Ethernet service across major business and data-centre areas. HKBN committed an additional HK$2 billion over five years to its core-network program; HKT's July announcement disclosed no project budget for the 3.2 Tbps route.
The capacity figures describe different layers: HKBN's 1.6 Tbps is an optical-core figure, its 100 Gbps is a customer-service figure, and HKT has not specified the customer interface behind its 3.2 Tbps headline. They demonstrate an alternative, not which network is faster, cheaper or more resilient.
An industry account also places HKT among telecom operators building AI-oriented optical and data-centre links. That broader competition further narrows HKT's “first” claim to the particular service and geography the company defined; it does not establish a unique business model.
HKT plans to use hollow-core fibre between Lok Ma Chau Loop and Tseung Kwan O. The company says it will deliver about 30% lower latency than traditional fibre and help keep thousands of GPUs synchronized, cutting compute time, cost and power consumption per bit.
The physical principle has experimental support, but the available research also shows why the scope of the claim is important. A 2020 Journal of Lightwave Technology experiment achieved error-free, single-lane 100 Gbps transmission over 100 metres of hollow-core fibre at a 2-micron wavelength. Against conventional solid-core fibre, the researchers measured a 30.95% physical-link latency reduction, equal to 1.5 microseconds per kilometre.
After adding a microsecond-scale switching network, the reported system-level reduction was 28% over one kilometre and 11% over 100 metres. These were experimental configurations, not tests of HKT's route. They show that fixed equipment delay can consume much of the propagation advantage on a short link.
HKT has not disclosed how much of its path will be hollow-core fibre, the conventional-fibre baseline, switching architecture or an end-to-end benchmark. Its 30% figure therefore cannot be translated into a 30% faster training run, inference response or scientific simulation from the evidence now available.
Sandy Ridge supplies the clearest potential reason to build northward, but it is on a different timetable from HKT's first phase.
Hong Kong's government awarded the Sandy Ridge site in March 2026 to Hong Kong Range Intelligent Computing Technology Company Limited on a 50-year land grant. The site exceeds 110,000 square metres. The tender premium was HK$581 million, and the successful bidder estimated HK$23.8 billion of cumulative investment from development through the first three years of operation.
The bidder expects to start operations within 42 months of the award, a window extending to roughly September 2029. The government also reported an expected 180,000 PFLOPS of computing power by 2032, equivalent to 36 times what it called Hong Kong's current capacity. That is a future project commitment, not installed compute today; the announcement provides no method for the PFLOPS estimate or workload basis. Computing throughput also does not by itself reveal how much inter-data-centre traffic the campus will generate.
Only one tender was received. That does not prove the project will be late, but it concentrates the delivery of this prospective demand source in one bidder. HKT's first route could be operating well before Sandy Ridge begins service, and HKT has announced readiness to extend there rather than a completed Sandy Ridge connection.
The end-of-2026 completion target is the first test. To evaluate the network rather than its label, customers and competitors will need HKT to disclose:
Until those facts arrive, HKT's announcement supports a narrower conclusion than its “superhighway” branding: the company intends to add a high-capacity data-centre route between Lok Ma Chau and Tseung Kwan O, but has not yet demonstrated the economics or application-level performance of a distributed AI cluster running across it.
Christian Faith Madison’s estate alleges GPT-4o sustained a months-long narrative of prophecy, sacrifice and resurrection before her death. Public safety disclosures and controlled research make that failure mode plausible, but neither establishes which model behavior she encountered or whether it caused her death.
Shanghai AI Laboratory has released Intern-S2-Preview-397B as downloadable weights and a hosted API, but its own 35B alternative offers the same context window and a far smaller disclosed scale while the larger model still lacks auditable comparative scores, pricing and physical-validation results.
Mixfont's free Decoy Font gives each glyph a sharp decoy and a blurred intended letter, creating a hurdle for pixel-based readers while leaving selectable text, known-image techniques and accessibility costs outside its protection.
Zhipu reportedly reached $1 billion in annual recurring revenue in July, roughly four times a March estimate, but the unconfirmed run rate is not annual sales and still sits far ahead of recognized cloud revenue while margins remain thin.
Cmpunlocker says patched Nvidia open kernel modules can expose 64GB of HBM on an 8GB CMP 170HX or 40GB on a 10GB card while removing an SM throttle. The exploit mechanism is documented, but the retained evidence does not independently validate the tool across full-memory workloads, error rates, power use or A100-class performance.
An unidentified industry source says at least one Nvidia board partner has physical RTX 50 Super cards but cannot sell them, with 3GB GDDR7 quoted at $60 to $70 per chip. The reported memory bill explains a plausible pricing problem, but neither Nvidia nor a second independent source has confirmed the products, the hold or the component terms.
Anthropic’s 20-year lease gives TeraWulf a customer for 401 megawatts of future AI infrastructure, but rent starts only after delivery and the proposed utility deal passes power and grid costs to TeraWulf, leaving financing and project margin unresolved.
OpenAI has put conversations and Projects back in its redesigned ChatGPT desktop app and enabled cloud Work threads to move across devices, correcting the launch's biggest usability failures without merging local Work or Codex histories.
Twenty-nine countries signed an agreement creating WAICO as an independent intergovernmental organization, while China paired the launch with capacity-building offers that are not yet confirmed as WAICO programs.
Meta reportedly plans to put departing AWS compute executive Dave Brown to work on its data-center buildout, adding hyperscale operating experience while leaving any customer-facing cloud business conditional and undefined.
Moonshot AI has made Kimi K3 available through its apps and API, pairing a 2.8-trillion-parameter architecture with early frontier-level results, but the model's open-weight claim cannot be tested until its weights and technical report arrive.
CIA Director John Ratcliffe said US intelligence is consistent with an estimate that Russian recruits last 20 to 30 minutes on Ukraine’s battlefield, but the public trail leads to an unsourced claim about assault troops and does not establish a representative average.
China has paired a five-year AI training offer for developing countries with cooperation centers, a weather-warning rollout and a new 29-country organization. The package gives Beijing a platform for influence, but no budget, selection rules or delivery timetable has been published.
The FastFlowLM team has joined AMD after building a Ryzen AI NPU inference flow on technology AMD already developed and distributed. The move could reduce model-support delays, but its financial significance, transaction perimeter and performance benefit remain undisclosed.
Federal intervention changed the launches of Anthropic's Claude Fable 5, Claude Mythos 5 and OpenAI's GPT-5.6, but through different legal and voluntary channels. The result is real government leverage over early customers without a common security threshold or durable review process.
A proposal developed with Treasury Secretary Scott Bessent would put an independent AI regulator under SEC oversight, but President Trump had not reviewed it and its tests, funding and enforcement authority remained unresolved.
San Francisco has demanded that Apple and Google cut off 13 apps capable of producing nonconsensual sexual deepfakes. The threatened case could test California’s new liability rules, but the confidential app list and the law’s “primary purpose” definition leave a central question unresolved.
Anthropic will keep Claude Fable 5 inside Max and Team Premium plans from July 20 at 50% of their usage limits, while Pro and Team Standard customers move to separately billed credits. The durable change is who gets bundled access, not a larger allowance for top-tier users.
Apple briefly topped Nvidia during Friday trading before Nvidia closed about $6 billion ahead. The filings show why that ranking is a weak AI verdict: Apple’s operating momentum predates its new Siri, while Nvidia’s rapid growth comes with concentrated customers and large forward commitments.
Anthropic has proposed paying Meta as much as $10 billion over two years for AI computing capacity, but early exit rights, undisclosed capacity and Meta's own reliance on outside infrastructure leave the economics—and the case for a durable Meta cloud business—unresolved.