Top Tech News Today, August 26, 2026: Amazon, Anthropic, Google, Microsoft, Waymo & More

It’s Wednesday, August 26, 2026, and the bill for the AI decade came due in cash, silicon, and human work. Meta is writing a check of up to $16.68 billion to settle claims it designed Facebook and Instagram to hook teenagers. Amazon is shutting Mechanical Turk after 21 years, ending the human-work marketplace that helped train the machine-learning era. OpenAI says its first custom inference chip now beats Nvidia’s Blackwell systems on AI work per watt in benchmark tests.

AI is no longer just a race to build the smartest model. It’s becoming a race for chips, memory, data centers, rockets, cloud distribution, and even the infrastructure governments depend on. China’s Moonshot AI is courting America’s biggest cloud companies, SpaceX is planning a staggering $100 billion launch complex, Samsung is rethinking how memory handles AI workloads, and cyberattacks are disrupting healthcare and government systems.

SoftBank is shopping a bond sale of up to $20 billion to keep funding that compute. SpaceX is putting $100 billion into Louisiana to make Starship launches an industrial-scale operation capable of lofting the next wave of satellites, including its planned orbital AI infrastructure. Bill Gates, meanwhile, wants some jobs declared off-limits to the systems the rest of the industry is racing to ship.

From AI and startups to regulation and Big Tech, here are the top tech news stories that explain how that collision is landing today.

Technology News Today

China’s Moonshot AI Seeks Landmark Kimi K3 Deals With Microsoft, Amazon, and Google

Chinese AI startup Moonshot AI is in early talks with Microsoft, Amazon, and Google over revenue-sharing agreements that could bring its Kimi K3 model to Azure, AWS, and Google Cloud. Moonshot is seeking as much as 30% of revenue generated from K3-related cloud services, according to Reuters. If completed, the arrangements would mark one of the most significant commercial integrations yet between a Chinese frontier AI developer and major U.S. cloud platforms. Negotiations are still working through issues including how revenue would be calculated, data access, and auditing model usage.

Kimi K3 is a 2.8-trillion-parameter model whose computing requirements make direct deployment impractical for many companies, making cloud distribution particularly important. The talks are also striking because Moonshot is facing growing scrutiny in Washington over alleged intellectual-property issues and access to advanced chips. The Alibaba-backed company has raised more than $2 billion and is also considering a Hong Kong listing. For Microsoft, Amazon, and Google, hosting K3 would give customers another high-end model choice while potentially exposing the cloud giants to new geopolitical and compliance questions.

Why It Matters: A commercial agreement would show that demand for competitive AI models can continue crossing U.S.-China technology barriers even as governments tighten controls around chips and AI.

Source: Reuters.

Meta Settles Teen Social Media Case for Up to $16.68 Billion and Rewrites Instagram and Facebook Defaults

Meta agreed to pay as much as $16.68 billion and overhaul how teenagers use Facebook and Instagram, ending a federal trial that had become the highest-profile test yet of claims that social platforms were designed to hook children. The proposed consent judgment, announced Wednesday by California Attorney General Rob Bonta and a bipartisan coalition of attorneys general, would impose a default two-hour daily cap for users under 18, block overnight use between midnight and 6 a.m. unless a parent lifts it, mute school-hour notifications, and add stronger age-assurance tools plus an independent auditor. About $12.7 billion would flow to participating states over ten years for youth online-safety programs; another $5.3 billion is contingent on YouTube and TikTok adopting similar teen limits. Meta denied wrongdoing and urged rivals to match the product changes.

The deal landed in the second week of trial in Oakland, before Instagram head Adam Mosseri was due back on the stand and before Mark Zuckerberg was expected to testify. States had alleged addictive design, misleading safety claims, and unlawful collection of children’s data, including leftover Cambridge Analytica-era privacy counts. For startups building social, creator, or teen-facing products, the settlement is a live blueprint: time limits, nighttime locks, and auditable age checks are moving from advocacy talking points into enforceable product requirements. Platforms that treat engagement maximization as the default now face a clearer legal and commercial cost for that choice.

Why It Matters: The largest U.S. youth-safety settlement yet turns teen time limits and age assurance from optional features into a compliance template the rest of the industry will be pressed to copy.

Source: TechStartups via CNBC.

DeepSeek Nears Massive Funding Round at $74 Billion Valuation Ahead of Possible 2027 IPO

Chinese AI startup DeepSeek is nearing completion of a funding round that would value the company at roughly 500 billion yuan, or $74 billion, before the new investment, according to people familiar with the talks. DeepSeek is seeking around 50 billion yuan, roughly $7 billion, with the round expected to close before the end of August. Existing investors, including Monolith and Shixiang Capital, are participating, and battery giant CATL is also among the backers cited in the report.

The financing could become a bridge to public markets. DeepSeek has begun preparing for a possible listing on Shanghai’s STAR Market and could file as early as the end of this year, with a market debut targeted for 2027. The financing has not yet closed, and the details come from unnamed sources discussing private negotiations, so valuation and investor composition may still change. Even with that caveat, the scale is notable. DeepSeek helped force a global reassessment of how efficiently competitive AI models could be developed, and a successful multibillion-dollar raise would give it substantially more resources for models, talent, infrastructure, and domestic chip partnerships.

Why It Matters: DeepSeek is moving from a disruptive AI research lab toward a heavily capitalized platform company, potentially giving China another major public-market AI champion.

Source: South China Morning Post.

SoftBank Weighs a $10–20 Billion Bond Sale to Refinance Its OpenAI Bet

SoftBank is discussing a $10 billion to $20 billion bond offering with banks to refinance part of the $40 billion bridge loan that underwrites its OpenAI investment, people familiar with the matter said. The notes could be sold in dollars and euros, possibly as early as September, and at the top of the range would be the largest Asian corporate bond deal of 2026. Proceeds would help repay the March 2026 bridge loan that matures in March 2027 and may also fund other AI investments. SoftBank still owes a final $10 billion OpenAI installment on October 1, which would take its total commitment to about $65 billion and an estimated 13 percent stake.

A spokesperson said it is reviewing various refinancing options and has not decided on anything. SoftBank is separately preparing a record ¥1 trillion retail bond in Japan and recently closed a $10 billion margin loan backed by the OpenAI stake. The financing stack is the hidden balance sheet of the AI boom: model companies need equity, and their largest backers need cheap debt to keep writing checks. Startups competing for the same limited pool of late-stage capital should assume OpenAI’s funding gravity now extends through the credit markets, not only venture funds.

Why It Matters: The AI capex cycle is migrating from venture term sheets onto the global bond market, which will price OpenAI risk for everyone else.

Source: The Straits Times.

Microsoft and Saudi Arabia’s HUMAIN Plan to Bring Arabic AI Models Into Foundry and Copilot

Saudi Arabia’s PIF-owned AI company HUMAIN has announced a long-term collaboration with Microsoft to bring its ALLAM Arabic-language models into Microsoft’s enterprise AI ecosystem. The companies plan to make ALLAM available through Microsoft Foundry, allowing developers and companies to build applications and AI agents around Arabic-language and regional business requirements. They also envision ALLAM-powered specialized agents working through Microsoft 365 Copilot.

The agreement is about more than distributing another model. HUMAIN engineers will work alongside Microsoft Forward Deployed Engineers to help organizations identify AI use cases, integrate systems into existing workflows, tune deployments, and move projects from prototypes into production. Importantly, the companies describe the Foundry and Copilot integrations as planned capabilities rather than products already generally available, and they have not announced a launch date. The partnership highlights Saudi Arabia’s broader strategy of building domestic AI capabilities while connecting them to global platforms. Arabic remains comparatively underserved by many leading enterprise AI systems, creating room for regionally trained models that better account for language, business context, and local deployment requirements.

Why It Matters: The Microsoft-HUMAIN deal shows the Middle East moving beyond buying AI infrastructure toward developing regional models that can plug directly into global enterprise software.

Source: Sharikat Mubasher.

Bill Gates Calls for “Human Reserved” Jobs as AI Enters a Turbulent Era

In a 6,000-word essay, Bill Gates argued that governments are unprepared for an AI transition he described as one of the most turbulent periods in human history. He proposed designating some work as “human reserved,” analogizing the idea to nature reserves: land that could be developed, but is left alone because the loss would be too great. He cited caregiving, delivering a terminal diagnosis, childcare, jury service, and parts of teaching and mental-health support as roles that should stay human even when a robot could do the mechanics. He also revived an older proposal to tax AI tokens and robots, and said new national and international institutions will be required because existing agencies were not built for a technology that moves this fast.

Gates told interviewers that an extreme version of the idea might initially reserve as much as 40 percent of jobs, and that some protections should be temporary for workers who cannot realistically retrain. He also said many executives are privately more alarmed than their public fundraising narratives suggest. The essay lands as OpenAI, Anthropic, and others race toward IPOs and as enterprises replace entry-level knowledge work with agents. For startups, the practical question is not whether Gates can legislate a reserve. It is whether labor rules, procurement standards, and public-sector buyers begin to hard-code human-in-the-loop requirements into healthcare, education, and government software.

Why It Matters: A mainstream industry founder is now arguing that some jobs should be legally or socially off-limits to automation, which would reshape product design and public contracts.

Source: The Guardian.

Anthropic Plans to Pitch Investors on an AI Market Worth More Than $30 Trillion

Anthropic is expected to tell prospective investors that the total addressable market for its AI technology exceeds $30 trillion, an extraordinary figure that could become central to its pitch ahead of a potential public offering. The estimate would exceed the $28.5 trillion opportunity SpaceX presented around its own record-breaking IPO, according to The Wall Street Journal. Anthropic is effectively arguing that AI models such as Claude could eventually address a large share of work performed across software, research, professional services, customer operations, coding, and other knowledge-intensive industries.

That number needs context. Total addressable market is a theoretical ceiling, not a revenue forecast, and assumes extremely broad AI adoption across economic activity. Still, the figure illustrates how AI companies are attempting to justify capital requirements and valuations once associated only with the largest public technology companies. Anthropic could reportedly seek more than $100 billion in an IPO at a valuation approaching $2 trillion. The more consequential question for investors is how much of the theoretical opportunity Anthropic can convert into recurring revenue while funding the immense computing infrastructure needed to serve increasingly capable models.

Why It Matters: Anthropic’s pitch signals that frontier AI companies increasingly see themselves as infrastructure providers for entire categories of human work, not conventional software vendors.

Source: The Wall Street Journal.

SpaceX Plans $100 Billion Louisiana Spaceport for Thousands of Starship Launches

SpaceX plans to invest about $100 billion in a massive new Louisiana spaceport that could eventually support thousands of Starship launches each year. The proposed Starbase Louisiana will occupy roughly 125,000 acres in Vermilion Parish on land formerly owned by Exxon, making it SpaceX’s fourth and largest launch complex. Construction is scheduled to begin in 2027, with the first Starship launch targeted for 2029. The company expects the site to become a major operating base for its reusable Starship system and future satellite deployments.

The scale points to a much larger change underway in the space economy. SpaceX is planning for launch cadence measured not in dozens of missions annually but potentially thousands, requiring industrial-scale propellant production, vehicle processing, transportation, and supporting infrastructure. Louisiana officials expect roughly 3,000 jobs over the next decade, while environmental groups are already calling for greater transparency around the development’s potential effect on coastal ecosystems. SpaceX says it will incorporate environmental preservation and restoration into the project. If Starship achieves reliable full reusability, facilities of this size could shift launch operations closer to an aviation-style model of frequent reuse.

Why It Matters: SpaceX is building physical infrastructure for a future in which reusable rockets fly at industrial scale, potentially reshaping satellite deployment, space logistics, and launch economics.

Source: Associated Press.

IBM Launches Granite 4.2 AI Models Built for Local Reasoning and Enterprise Agents

IBM has released Granite 4.2, a new family of open-weight AI models aimed at enterprises that want reasoning and agent capabilities without sending every workload to a giant cloud-hosted model. Granite 4.2 comes in 3-billion, 8-billion, and 30-billion-parameter versions and supports a native 128,000-token context window. The larger 8B and 30B models received specialized reinforcement learning for agentic tasks such as using tools, working in terminals, searching for information, and carrying out multistep instructions.

The release reflects a growing split in the AI market. Frontier Labs continues racing toward ever-larger models, while enterprises increasingly want smaller systems that run locally, stay under company control, and offer predictable costs. IBM is positioning Granite around that second market. The models include reasoning modes, native tool calling, coding support, and an Apache 2.0 license, making them available for commercial use and customization. They are also distributed through platforms including Hugging Face, Ollama, GitHub, LM Studio, and several inference providers. For companies building internal agents around sensitive corporate data, model size and deployment flexibility can matter as much as benchmark leadership.

Why It Matters: Granite 4.2 shows enterprise AI shifting toward smaller, controllable reasoning models that companies can deploy on their own infrastructure, rather than relying exclusively on frontier-model APIs.

Source: Ars Technica.

Samsung Puts AI Processing Inside Memory With LPDDR5X-PIM

Samsung is moving AI computation closer to the data itself with LPDDR5X-PIM, a new memory design that places processing logic alongside DRAM cells. Detailed at Hot Chips 2026, the technology allows certain calculations to happen inside memory instead of repeatedly moving data between memory and a separate processor. Samsung says that approach can dramatically reduce a major AI inference bottleneck. In preliminary tests, LPDDR5X-PIM delivered 2.28 times faster model runtime and 3.01 times greater token throughput than conventional LPDDR5X.

Peak bandwidth rose from 76.8 GB/s on LPDDR5X-9600 to a theoretical 614 GB/s in PIM mode, an eightfold increase. Samsung tested the design with an 8-billion-parameter Llama 3.1 model on an edge AI accelerator. The company acknowledged that optimization work, including accuracy tuning, remains underway, so the benchmark results should be treated as preliminary rather than production guarantees. The broader idea is important because high-bandwidth memory has become one of the most expensive and constrained components in AI hardware. Bringing useful computation into lower-power LPDDR memory could open another route for AI inference in laptops, edge systems, and other devices where HBM is impractical.

Why It Matters: Processing-in-memory could attack one of AI hardware’s biggest inefficiencies: constantly moving enormous amounts of data between memory and processors.

Source: Tom’s Hardware.

OpenAI’s Data Center Chief Leaves as the AI Company Ramps Up Compute Expansion

Chris Malone, OpenAI’s head of data centers, has left the company as it continues reorganizing the team responsible for securing the enormous computing capacity behind ChatGPT and its future models. Malone joined OpenAI in March 2025 shortly after the company announced Stargate, its ambitious infrastructure initiative with Oracle and SoftBank. His departure was reported Wednesday, as OpenAI prepares for a possible public offering while spending heavily on computing infrastructure.

The personnel changes matter because infrastructure has become one of OpenAI’s most strategically important functions. Training frontier models requires enormous clusters, but inference demand from ChatGPT, coding products, APIs, and autonomous agents creates an equally difficult capacity-planning problem. OpenAI has increasingly diversified how it obtains that compute, combining cloud partnerships with leases and direct involvement in large data-center developments. Malone’s exit follows several other senior management changes and comes while the company reshapes its infrastructure leadership. Executive turnover alone does not indicate that OpenAI’s buildout is slowing, but continuity matters when projects span utilities, chips, networking, real estate, financing, and multi-year construction schedules.

Why It Matters: For frontier AI companies, securing electricity, chips, data centers, and network capacity is now as strategically important as developing the models themselves.

Source: The Wall Street Journal.

Amazon Will Shut Mechanical Turk on September 30 After 21 Years of Crowdsourced Work

Amazon told workers and requesters that AWS Mechanical Turk will close on September 30, 2026, ending the marketplace Jeff Bezos once branded “artificial artificial intelligence.” The 2005 service farmed out Human Intelligence Tasks—image labeling, transcription, surveys, content moderation—to more than 500,000 people at a few cents a task. New customer sign-ups already stopped on July 30, the same day Amazon closed SageMaker Ground Truth and Amazon Augmented AI to new users. The company offered little explanation beyond a routine “assessment” of programs and services, and pointed remaining customers to a FAQ on refunds and payout settings.

The shutdown marks how far model training has moved. Specialized labeling firms such as Scale AI, Mercor, and Prolific absorbed the high-value annotation work that once flowed through MTurk, while large language models now handle many of the cheap classification jobs that made the platform famous. Quality had already become a problem: academic work in 2023 found a large share of MTurk workers using AI to complete tasks meant to train AI. For startups that still treat Mechanical Turk as a default data pipe, five weeks is a short runway to migrate pipelines, re-price human review, and decide whether synthetic data or contracted annotators can replace a two-decade commodity labor market.

Why It Matters: The original crowdsourcing layer of the modern AI stack is disappearing, forcing data and evaluation startups to compete on quality and contracts rather than on a public penny-task exchange.

Source: TechStartups via Amazon, CNBC.

Boston Scientific Cyberattack Disrupts Global Operations and Customer Shipments

Medical-device giant Boston Scientific disclosed Wednesday that a cybersecurity incident has disrupted parts of its global operations, including systems used to process and ship customer orders. The company detected the incident on August 25 and said affected systems include both information technology infrastructure and business applications. Boston Scientific has brought in outside cybersecurity specialists and is working to restore operations, but it has not provided a timetable for full recovery.

The company has not yet said whether sensitive customer or patient information was accessed, nor has it determined whether the attack is likely to have a material financial impact. That distinction is important: the confirmed problem so far is operational disruption, not a confirmed data breach. Still, interruption of order processing at a major medical-device manufacturer demonstrates how cyberattacks can move beyond stolen records and directly affect physical supply chains. Boston Scientific joins a growing list of healthcare and life-sciences companies dealing with cyber incidents this year, increasing pressure on the sector to treat manufacturing, logistics, and business applications as part of the same security perimeter as clinical and patient systems.

Why It Matters: Cyberattacks against medical technology companies can disrupt the movement of real-world healthcare products even when the full extent of data exposure remains unknown.

Source: Fierce Biotech.

Hackers Are Actively Exploiting Critical Gitea Flaw to Run Commands on Servers

Attackers are exploiting a critical vulnerability in Gitea, the self-hosted Git platform used by software-development teams to manage repositories and DevOps workflows. The flaw, tracked as CVE-2026-60004, lets users with repository write access execute arbitrary shell commands under the Gitea service account by abusing the platform’s diffpatch API. The risk is higher on default installations because Gitea enables self-registration, potentially letting an outside attacker create an account and repository before triggering the exploit.

Gitea fixed the vulnerability in version 1.27.1 on July 27, but active exploitation has now pushed the issue into a more urgent phase. CISA added the flaw to its Known Exploited Vulnerabilities catalog and ordered U.S. federal civilian agencies to remediate affected servers by August 28. Shadowserver currently tracks nearly 5,000 internet-exposed Gitea instances, although that figure includes systems that may already be patched or used as honeypots. Reports cited by BleepingComputer indicate attackers have used vulnerable servers to install cryptocurrency-mining malware. Because source-code repositories often sit close to deployment credentials and software pipelines, compromise can carry consequences beyond the Gitea server itself.

Why It Matters: An actively exploited vulnerability in a source-code platform can become a software-supply-chain problem if attackers gain access to build systems, secrets, or production workflows.

Source: BleepingComputer.

Pro-Russian Hackers Claim Norway’s Largest Government Digital-Service DDoS Attack

A pro-Russian hacking group called Server Killers has claimed responsibility for a sustained denial-of-service campaign targeting Norway’s government digital infrastructure. The attack has continued for several days and affected systems operated by the Norwegian Digitalization Agency, including infrastructure that gives citizens a single login across multiple public services. An agency spokesperson described it as the largest attack its systems have faced, though officials kept services available for most of the incident.

Server Killers said the campaign was retaliation for Norway’s renewed security cooperation with Ukraine, including deeper collaboration around technologies such as drones. Norwegian authorities have not independently confirmed that the group was responsible, so the attribution remains a claim rather than a verified conclusion. The incident nevertheless illustrates a persistent pattern across Europe: politically motivated groups using relatively inexpensive cyber operations to pressure governments and create public disruption around geopolitical disputes. DDoS attacks generally do not require penetrating internal government networks, but large enough campaigns can still interrupt citizen-facing systems and force governments to divert significant resources to defense and mitigation.

Why It Matters: Public digital infrastructure has become another frontline for geopolitical pressure, giving hacktivist groups a way to disrupt government services without conventional military capabilities.

Source: Associated Press.

Slovakia Moves to Ban Social Media Accounts for Children Under 16

Slovakia’s government approved legislation Wednesday that would prohibit children under 16 from creating social-media accounts, putting the country among a growing number of governments seeking stricter age controls on online platforms. The proposal would require users to prove their age when opening accounts, while attempting to prevent platforms from receiving unrelated personal information during verification. The bill still needs parliamentary approval to become law.

The government says the proposal is aimed at reducing risks including addictive behavior, social pressure, manipulation, disinformation, extremist content, and radicalization among younger users. The proposed age-verification framework would be tied to European Union digital-services rules, making implementation especially relevant for platforms operating across the bloc. The larger challenge is both technical and political: governments want reliable proof that a user is old enough to access a service without creating new databases of identity documents or undermining privacy. As more countries impose minimum ages, major platforms may eventually need standardized age-assurance systems rather than separate compliance mechanisms for every jurisdiction.

Why It Matters: Social-media regulation is moving from parental-control tools toward mandatory age assurance, potentially forcing major platforms to rethink account creation globally.

Source: Reuters.

London Robotaxi Launch Slips as Waymo, Wayve, and Baidu Wait on Regulatory Guidance

London’s push to launch fully driverless robotaxis is running behind schedule, with Waymo, Wayve, and Baidu now unlikely to begin full commercial services in the city this year. Transport for London has yet to issue key guidance operators need to navigate the approval process, while technical and regulatory requirements remain unresolved. Pilot programs using human safety drivers are still expected, but no vehicle has yet received the necessary approval from the Driver and Vehicle Standards Agency for fully driverless commercial operation.

The delay shows why autonomous-driving deployment remains as much a regulatory challenge as an AI problem. Companies can demonstrate increasingly sophisticated vehicles, but commercial service requires governments to settle questions around certification, liability, local operating rules, emergency response, and public safety. London has attracted several of the most advanced U.S., Chinese, and British autonomous-driving companies, making it an important test of whether one regulatory framework can accommodate competing technical approaches. The UK government still expects broader self-driving deployments beginning in 2027 after testing and local approvals. A slower launch does not necessarily mean the technology has failed, but it pushes revenue farther out and gives other cities more time to establish themselves as early autonomous-mobility markets.

Why It Matters: Robotaxi commercialization increasingly depends on regulatory execution, not just better autonomous-driving software.

Source: Financial Times.

Hearing Tech Startup Legato Emerges With $12 Million and AI-Powered Smart Glasses

Legato has emerged from stealth with $12 million in funding and plans to launch smart glasses that use AI to help people with mild to moderate hearing loss. The startup’s Legato Frames integrate hearing-assistance components into the arms of ordinary-looking eyewear, so users don’t need to wear traditional hearing aids. Legato was founded by Mehul Trivedi and Steve Romine, two executives with deep experience in smart glasses and hearing technology from Bose, EssilorLuxottica, and Audicus.

The glasses use an AI-based audio system designed to distinguish voices from background noise and amplify speech, addressing one of the hardest environments for hearing assistance: busy restaurants, meetings, and other noisy spaces. Legato says its dual-speaker design directs sound toward the wearer while reducing sound leakage. The company expects to launch the frames later this fall through selected eye-care providers. Funding from Neotribe Ventures, Listen, and Village Global has primarily supported product development. Pricing has not been disclosed, although Legato says the device should cost substantially less than many traditional hearing-aid solutions. As with any new assistive-health product, real-world performance will ultimately matter more than early technical claims.

Why It Matters: Legato represents a practical direction for AI wearables: embedding intelligence into familiar products to solve a specific everyday problem rather than creating another general-purpose gadget.

Source: TechCrunch.

Huawei and HP Sign Global Wi-Fi Patent Cross-Licensing Agreement

Huawei and HP have signed a multiyear global agreement allowing the two technology companies to cross-license Wi-Fi patents. The arrangement gives HP access to specified Huawei Wi-Fi technologies while providing Huawei reciprocal rights to patents held by HP. Neither company disclosed financial terms, and the agreement is focused on intellectual-property licensing rather than a new hardware partnership or broader strategic alliance.

The deal offers a useful reminder that the technology supply chain remains deeply interconnected, even as geopolitical pressure encourages companies and governments to reduce cross-border dependencies. Wi-Fi is built around standardized technologies implemented by manufacturers worldwide, so companies holding key patents can remain commercially relevant far beyond their domestic markets. Huawei has invested heavily in communications research and increasingly uses licensing to monetize parts of that portfolio. HP, meanwhile, needs access to standardized technologies across PCs, printers, and connected devices. Cross-licensing agreements can reduce litigation risk while giving both sides greater freedom to develop products without repeatedly negotiating around overlapping intellectual property.

Why It Matters: The Huawei-HP agreement illustrates how global technology standards can preserve commercial cooperation even when broader U.S.-China technology relations remain strained.

Source: TechNode.