September 05, 2026: Week Ends Better Than It Started

Last week, Dow Jones fell 0.3%, S&P 500 was flattish, and Nasdaq composite gained 4%. Nvidia agreed to buy Hugging Face for US$12.9 billion (US$11.9 billion plus a US$1 billion retention pool), giving it the main distribution channel for open-source models. Disclosures show Nvidia now carries roughly US$230 billion of lease backstops and residual-value support, including US$105 billion on an OpenAI Ohio lease, on top of US$72.5 billion of equity stakes, making it “lender of last resort” to the AI buildout. Big Tech booked over US$160 billion of “other income” last quarter on AI stake gains, double the prior quarter’s ~US$69 billion, largely from SpaceX’s listing (post-xAI merger). Anthropic marks at Alphabet and Amazon drove record pre-tax profit. Broadcom revenue rose 86% to US$29.6 billion with AI chips up 221% to US$16.7 billion. The company guided AI revenue to US$115 billion in 2027 with a line of sight to US$230 billion in 2028 and EPS above US$30. Dell revenue rose 58% to US$47 billion, FY guidance up US$25 billion to US$192 billion on US$74 billion of AI servers. HPE raised guidance but fell on memory-led supply constraints. Snowflake product revenue grew 37% to US$1.49 billion and shares rose 20%. Palo Alto grew 34% with XSIAM ARR at US$700 million. Microsoft disclosed Azure revenue of US$101.9 billion, +40%. OpenAI shipped GPT-6 Astra at US$10/US$50 per million tokens, hinting at AGI. Its ad business now reportedly run-rates US$1 billion ARR versus projected US$2.4 billion. Oura filed to IPO on US$1.2 billion revenue (+74%) and US$262 million FCF. SB Energy, Moonshot (~US$50 billion) and Enflame also filed to go public. Shein debuted in its IPO flat at US$26 billion. FTC sued Amazon over ad auctions, while Google avoided an ad-tech breakup. In Sophic client news, Cybeats posted Q2 revenue of $815,769 (+9.5%), a narrowed $764,074 loss, $261,643 cash plus a $1.9 million placement. The Company also expanded a six-figure medical-device contract by 76% with its first RAVEN’s commercial rollout. The Company has 8 active POCs. Hybrid Power Solutions filed a patent on an AI-driven hybrid energy management system.

Sophic Client Cybeats Technologies Corp. (CYBT-CSE,CYBCF-OTCQB) announces second quarter Fiscal 2026 financial results.

Cybeats announced its financial results for the three and six months ended June 30, 2026. Financial Highlights for the three months ended June 30, 2026 (“Q2 2026”) with comparatives for the three months ended June 20, 2025 (“Q2 2025”): Revenue in Q2 2026 was $815,769, an increase of 9.5% or $71,105 over Q2 2025 revenue of $744,664. This growth was due to the addition of new customers and the expansion of our existing customers. Net loss in Q2 2026 was $764,074 versus a loss of $792,922 in Q2 2025, an improvement of 4% or $28,848. Cash expenses decreased slightly to $1,535,904 in Q2 2026 compared to $1,536,727 in Q2 2025. Cash at the end of the quarter was $261,643 compared to $1,072,900 at the end of Q1 2026. Subsequent to quarter end, the Company closed an oversubscribed non-brokered private placement for $1.9 million. Management Commentary: “We are seeing the business start to accelerate, and the opportunities in front of us today are larger and broader than they have ever been,” said Justin Leger, CEO of Cybeats. “We are winning new enterprise customers, expanding into new markets, and most importantly, seeing our channel partners begin to convert opportunities into commercial business. This is exactly what we wanted to see as we built out our partner-led strategy: broader market reach, access to new customers and geographies, and another meaningful engine for growth alongside our direct sales efforts to scale efficiently. With demand building across regulated industries, a growing global pipeline, and several opportunities continuing to advance through the sales process, we believe Cybeats is very well positioned for a strong second half of the year.” Business Outlook: Demand for Cybeats’ software supply chain security solutions remains strong, supported by increasing regulatory requirements, enterprise adoption, and the growing complexity and velocity of modern software development. The accelerating adoption of AI-assisted software development is further contributing to this trend, as organizations increasingly rely on AI tools to generate and modify code, introduce third-party components, and accelerate software release cycles. Management believes these practices are increasing the importance of software transparency, component traceability, and automated software supply chain risk management, creating an additional tailwind for SBOM adoption. During and subsequent to the quarter, the Company added new enterprise business and continued to advance opportunities across medical devices, industrial automation, defence, healthcare, telecom, automotive, and critical infrastructure. Cybeats is also seeing growing traction from its channel and OEM strategy, with partner-generated opportunities progressing through the sales cycle and converting into commercial engagements. As of August 14, 2026, the Company had 8 active POCs and a further 6 scheduled for Q3 and Q4, providing a growing pipeline of potential recurring revenue opportunities across direct enterprise, existing customer expansion, and partner-led channels. Management believes Cybeats enters the second half of 2026 with the strongest commercial opportunity set in its history, supported by increasing partner activity, a growing ARR base, and continued regulatory and geopolitical tailwinds. Governments around the world are placing increasing emphasis on the security and resilience of critical national infrastructure and the software supply chains that underpin it, while regulatory frameworks are increasingly translating these priorities into concrete cybersecurity, vulnerability management, software transparency, and incident reporting requirements. These trends are creating sustained demand for technologies that provide organizations with greater visibility and control over their software supply chains, including SBOM management. Regulatory momentum includes the EU Cyber Resilience Act, with vulnerability and incident reporting obligations beginning September 11, 2026. https://t.co/x1bW754B7F

Sophic Client Cybeats Technologies Corp. (CYBT-CSE,CYBCF-OTCQB) expands contract with global medical device manufacturer by 76%.

Cybeats announced a 76% revenue expansion of its existing six-figure commercial engagement with a global medical device manufacturer for SBOM Studio. The expansion reflects the continued adoption of Cybeats’ platform across the customer’s software security program. The renewed engagement also includes the onboarding of RAVEN, Cybeats’ new agentic AI-native intelligence layer, marking the first commercial rollout of RAVEN with a customer. Consistent with industry practice, the customer has requested to remain anonymous due to internal cybersecurity policies and procurement confidentiality. This is common among organizations operating in highly regulated industries, where cybersecurity technologies and vendors are often not publicly disclosed. Cybeats also continues to build its future revenue pipeline, with eight active proof-of-concept (POC) engagements currently underway across multiple sales channels. These POCs span regulated industries including healthcare, industrial, and critical infrastructure, supporting both new customer acquisition and future expansion opportunities. “Strong enterprise software businesses are built by delivering measurable customer value over time,” said Justin Leger, CEO of Cybeats. “This expansion demonstrates the confidence our customers have in SBOM Studio and the results they are achieving. This expansion represents an increase in revenue, scope, and contract duration by an important customer. Combined with our strong and growing Net Revenue Retention and an expanding pipeline of active proof-of-concept engagements, we believe these metrics reflect continued momentum in our commercial execution and validate our land-and-expand strategy.” https://t.co/tz062FCh33

Sophic Client Hybrid Power Solutions (HPSS-CSE, HPSIF-OTC) files Patent application for Advanced AI-Powered Energy Management System.

Hybrid Power Solutions is pleased to announce the filing of a comprehensive patent application for a groundbreaking AI-driven energy management system. This innovation is designed to optimize hybrid power deployments integrating battery storage, renewables, generators, and grid connections for industrial, remote, and mission-critical applications. The patent application details a sophisticated system featuring: A communication interface supporting multiple protocols including Modbus and CANbus. A sensor interface integrating GPS, accelerometers, humidity, Hall effect current sensors, voltage, acoustic, Bluetooth, and Wi-Fi capabilities. A machine learning engine that analyzes power consumption and generation patterns to generate autonomous control actions, continuously learning from operational data to optimize efficiency without user intervention. Predictive maintenance capabilities using feature extraction, degradation trend analysis, and remaining useful life estimation. Dynamic weighting for optimization objectives, contextual adjustments (e.g., electricity prices, carbon intensity), and confidence scoring for AI decisions with human review triggers. Renewable energy management with advanced MPPT, wind turbine control, microgrid coordination, and unified management of solar, wind, battery storage, and controllable loads. Extreme weather preparation, power quality monitoring per IEEE standards, fault compensation, simulation modeling, and more. Francois Byrne, CEO of Hybrid Power Solutions, commented: “This patent application represents a major leap forward in intelligent hybrid power systems. By embedding AI, machine learning, predictive analytics, and real-time optimization, we are enabling our customers to achieve superior efficiency, lower operating costs, reduced emissions, and enhanced reliability—especially in harsh or remote environments where our Spark, Batt Pack, and Solar Tarp solutions are already making an impact. This technology will further differentiate Hybrid in the market and support our transition toward smarter, more autonomous clean power platforms. This could be a new source of recurring revenue once implemented across the product line.” The system aligns closely with Hybrid’s ongoing development of its embedded IoT and unified fleet management platform, which incorporates predictive analytics and machine learning for remote monitoring and optimization. It also complements tools like the Company’s new customer-facing power usage calculator, which helps users evaluate hybrid solutions, estimate ROI, cost savings, payback periods, and CO₂ reductions based on real-world data. This filing underscores Hybrid’s commitment to innovation in portable and hybrid energy storage, building on prior intellectual property achievements and positioning the Company for expanded opportunities in construction, mining, defense, First Nations microgrids, and other high-demand sectors. https://tinyurl.com/f6wk34uf

Oura files to go public.

Oura Inc., the company that sells rings that track people’s health and activities, has turned into an enormously profitable enterprise. Its IPO filing, disclosed on Thursday, showed that it generated US$262 million in free cash flow in the nine months ending June 30. Oura is also growing quickly, lifting revenue 74% to US$1.2 billion in the same period. The number of rings sold hit 3.1 million in the nine month period, up from 1.8 million a year earlier, while the number of paying members doubled to five million. Paying members get extra benefits. Oura is one of the highest profile examples of wearables startups, cashing in on consumers’ desire to track their health in detail. Oura is competing with Apple’s watch, Google’s Fitbit as well as other standalone firms such as Whoop. https://tinyurl.com/49cwv5v2

SoftBank’s SB Energy files to go public.

SoftBank’s affiliate company SB Energy filed to go public, laying out in detail its ambition to become a neocloud. So far, however, as the filing showed, SB Energy has “no data center capacity” currently operating. SB Energy currently generates revenue from the sale of electricity to power authorities, energy service companies and others, with power coming from solar and batteries. Revenue last year totaled US$213.5 million and the company burned more than US$1 billion, mostly in capex. Nvidia is investing US$3 billion in SB Energy: US$1.5 billion via a prepaid forward contract with SB Energy parent Energy Global and US$1.5 billion at the time of the IPO. The company is building data centers for customers including SB Energy’s controlling shareholder, SoftBank, and OpenAI, which SoftBank is also a big shareholder in. Nvidia is backing the development of one of the data centers being built for OpenAI. https://tinyurl.com/dw5v3bas

Moonshot AI files confidentially for Hong Kong IPO.

Moonshot AI, whose Kimi K3 open-source model has become a global hit, has filed confidentially an application for a Hong Kong initial public offering, Chinese media outlet LatePost reported Wednesday. Moonshot’s path to its IPO hasn’t been easy. The company went through an overhaul to unwind its original corporate structure with an overseas holding company, after Chinese regulators’ new stance made it harder for companies with such a structure to go public. Moonshot’s listing is subject to Chinese regulatory approval. The company is currently in talks with investors to raise a new pre-IPO funding round at a valuation of US$50 billion, according to a person with knowledge of the round. The company’s revenue is growing rapidly thanks to the popularity of Kimi K3 among enterprise customers and developers. Kimi K3, released in July, surprised the world with its frontier-level coding and agentic capabilities. Some of Moonshot’s domestic competitors are already listed. Z.ai and MiniMax, which both went public in Hong Kong in January, were valued at US$6.7 billion and US$6.5 billion, respectively, at the time of their IPOs that each raised hundreds of millions of dollars. Z.ai’s current market capitalization is about US$66 billion, while MiniMax’s is around US$15 billion. https://tinyurl.com/5b73r2ma

Tencent-backed China chipmaker Enflame seeks US$911 million in IPO.

Shanghai Enflame Technology Co., backed by Tencent Holdings Ltd., is seeking to raise about 6.12 billion yuan (US$911 million) in an initial public offering, as the last of the so-called “four little dragons” of leading Chinese artificial intelligence chipmakers taps the public market. The company, in which Tencent holds a 20% stake and remains a dominant customer, priced its offering on Shanghai’s technology-heavy STAR Market at 142.18 yuan per share, according to an exchange filing Monday. The sale will consist of about 43 million shares, or 10% of its enlarged share capital following the IPO. https://tinyurl.com/bdfvbvpx

Shein ends flat in debut after raising US$1.7 billion in Hong Kong IPO.

Shares of Chinese online fashion giant Shein ended flat on its first day of trading on the Hong Kong stock exchange, capping off the company’s convoluted multiyear quest to go public. Shein’s shares opened at HK$48.56 on Tuesday, and dropped as much as 10% before shedding the losses and closing at HK$48.5, according to Hong Kong stock exchange data. By comparison, the benchmark Hang Seng Index dipped 0.93%. As of Tuesday market close, Shein had a market capitalization of US$26 billion, a far cry from the height of US$100 billion reached during a private funding round in 2022. The company raised HK$13.6 billion (US$1.7 billion) in the initial public offering. Shein for years tried to downplay its China roots and position itself as a global company, as its fashionable, ultracheap clothing gained popularity among consumers in the West. The company tried to go public in both New York and London, but the attempts were thwarted by criticism against its labor and environment practices and China’s straining ties with the U.S. Wearing a white, Shein-branded T-shirt like other employees, Sky Xu, the company’s elusive founder and CEO, attended the listing ceremony in Hong Kong but stayed out of the limelight most of the time. He didn’t deliver a speech or participate in the iconic Hong Kong-style gong-striking ritual on stage to mark the beginning of trading. Shein’s employees struck the gong and CFO Leigh Gui delivered a short speech. https://tinyurl.com/58xh97z9

Crusoe reportedly raises US$3 billion at a US$30 billion valuation.

Data center developer Crusoe, which counts Meta, Microsoft, and OpenAI as its customers, has raised a new US$3 billion round at a US$30 billion valuation, Bloomberg reported. The deal is being co-led by Atreides Management and Valor Equity Partners, and includes participation from Mubadala Capital, the asset management subsidiary of Abu Dhabi’s sovereign wealth fund Mubadala. The company recently signed a massive US$13 billion, five-year cloud contract to supply quantitative trading firm Jane Street with GPUs and AI infrastructure, Bloomberg reported. The fresh fundraise comes 10 months after Crusoe raised a US$1.38 billion round at a US$10 billion valuation last October. Launched in 2018 as a crypto mining operation powered by flared natural gas, Crusoe has since pivoted into a major AI infrastructure and cloud provider that is best known for developing hyperscale data center campuses for clients like Oracle and OpenAI. The company recently met with investment bankers, including Goldman Sachs and Morgan Stanley, to discuss a potential near-term IPO, Axios reported last month. https://tinyurl.com/mu5wt8jp

Chinese AI model firm z.ai’s API sales surge in first half.

Chinese AI firm Z.ai’s revenue in the first half jumped nearly fivefold from a year earlier to 954 million yuan (US$142 million), boosted by skyrocketing sales of cloud-based AI model services. Z.ai, whose stock is listed in Hong Kong, said its “open platform and API revenue,” which includes the sales of its models through application programming interfaces and other services such as coding plans, surged more than 28 times to 825 million yuan. Its loss for the period narrowed 12% to 2.07 billion yuan. The fast growth in API sales in the first half reflects the popularity of the company’s GLM-5.1 and GLM-5.2 models, which delivered strong coding capabilities at affordable prices. Last month, Z.ai released its new-generation model, GLM-5.3, which is challenging U.S. frontier models in coding and agentic tasks. But Z.ai is also leading an intensifying price war among Chinese AI models, which could affect its revenue and margins in the second half of this year. Last week, the company launched GLM-5.3 Flash, a low-cost model that only charges $0.15 for input and $0.50 for output per 1 million tokens. https://tinyurl.com/4kh4h376

Anthropic said to reach US$35 billion compute deal with Nvidia-backed Lambda.

Anthropic signed a deal to rent US$35 billion worth of compute capacity from Nvidia-backed cloud provider Lambda, The Wall Street Journal reported Monday. Nvidia, an investor and supplier to Lambda, will supply its chips for the data center providing the capacity to Anthropic and hold the lease on it, the report said. Hut 8, a former bitcoin mining company, is developing the data center in Nueces County, Texas. Anthropic has been making efforts to alleviate its compute crunch by signing a series of agreements that give it access to servers that run AI. The Lambda deal comes less than a week after Anthropic secured $45 billion of compute from U.K. cloud provider Nscale. Earlier this year, it agreed to pay SpaceX US$40 billion for compute through 2029. https://tinyurl.com/5buxhu5z

Nvidia has become a banker to the AI boom, putting it on dangerous ground.

Nvidia recently provided backstops that could put it on the hook for some US$230 billion in lease obligations and residual-value deals, including a US$105 billion backstop for an OpenAI lease in Ohio and potentially up to US$125 billion in “residual-value support” for financing deals with Wall Street heavyweights. That support would likely entail Nvidia’s assuring that the value of assets backing loans won’t fall below a predetermined level. Nvidia also recently signed deals with two young cloud-computing companies in Australia where it acts as a customer of last resort if they can’t find others to rent their computing power to. In exchange, Nvidia gets a share of the companies’ revenue from those assets above an agreed-upon level. The backstops give lenders confidence to extend financing with which the companies buy Nvidia’s chips. Those promises don’t include Nvidia’s agreement last year to buy unsold cloud-computing capacity from CoreWeave as part of a US$6.3 billion deal. Nor do they include the equity stakes Nvidia has taken in many of the cloud companies and AI labs that it is providing backstops for, including CoreWeave. The company had US$72.5 billion of public and private-company stakes at the end of its last fiscal quarter. Part of what Nvidia is doing is spreading its bets so it is less reliant on a handful of tech giants like Amazon.com and Microsoft. Those companies are running out of financial steam. Yields on AI-linked Big Tech bonds are surging because there is such a large supply of them. With no end in sight for debt issuances, borrowing costs for AI projects will likely stay high. And the tech companies are already heavily committed to data center projects, with US$904 billion of leases not yet started and US$1.52 trillion of purchase commitments largely related to AI, per their most recent financial statements. There is thus some logic in Nvidia trying to generate sales from smaller and less-creditworthy companies. But history also provides examples of companies for whom courting less-creditworthy customers backfired spectacularly. https://tinyurl.com/mws6mnbb

Nvidia to buy Hugging Face for US$12.9 billion.

Nvidia has agreed to buy AI model platform Hugging Face for US$12.9 billion, the latest step by the chip giant to use its financial heft to exert greater control in the AI ecosystem. Nvidia’s announcement confirms The Information’s report last week that Nvidia and Hugging Face had agreed to a deal for US$12.9 billion. The acquisition of Hugging Face puts Nvidia in charge of a key distribution channel for millions of models in the race among open-source model developers to catch up to closed-source AI models from Anthropic and OpenAI. Hugging Face will remain an open platform, Nvidia said in a press release. Nvidia is paying US$11.9 billion and setting up a US$1 billion equity retention program for Hugging Face employees who join Nvidia, according to a securities filing. The acquisition is the latest among a flurry of investments and license-and-hire deals Nvidia has struck in recent months. Nvidia has put tens of billions of dollars in everything from AI application developers such as Perplexity and Poolside to companies that develop land and power for data centers. Nvidia has been boosting its own effort to develop Nemotron-branded open-source AI models and has committed to spend tens of billions developing its own open models. https://tinyurl.com/47d8wemv

GoPro to sell majority stake for US$285 million, pivot to AI and Defense. Action camera company GoPro has agreed to sell a majority stake to optics manufacturer Starman Optical, the two companies said Tuesday. Starman Optical will buy 90% of GoPro’s shares for US$285 million in cash, or US$1.14 a share. GoPro shares jumped 40% on Tuesday. The deal is expected to close by the end of this year and will help GoPro, which will remain a public company, clear its US$92 million in debt, the companies said. GoPro cameras designed to record videos of actions like skiing or surfing had propelled its shares to as high as US$93 in 2014. But shares tumbled in subsequent years on declining sales and, more recently, increased memory costs. This June, GoPro expressed “substantial doubt about the Company’s ability to continue as a going concern.” The merger would position GoPro to move into “areas of national security related to cameras, optics and AI infrastructure,” GoPro founder and CEO Nicholas Woodman said in a statement. The combined company hopes to target its optics and imaging technologies towards the defense, datacenter, aerospace and robotics markets. https://tinyurl.com/5aa9a55y

Big Tech profits get US$160 billion boost from gains on stakes in other AI companies.

Big tech giants booked a more than US$160 billion windfall last quarter from investments in other AI companies, flattering their earnings and raising concerns that paper gains are overstating the strength of the AI boom. Alphabet, Amazon, Nvidia and Microsoft all reported substantial boosts to their pre-tax profits in their most recent results from the ‘other income’ line in their accounts, which largely comes from valuation gains on equity stakes in other leading AI companies. These one-off valuation boosts, derived in large part from enthusiasm around AI, risk distorting the financial picture at a time when investors are closely scrutinising tech earnings to gauge the health of the AI ecosystem. Big Tech groups’ investments in OpenAI and Anthropic have increasingly shown up in their financial reporting in recent quarters, as accounting rules dictate that changes in the value of equity investments at the end of each quarter appear as part of the company’s profit or loss. But their equity gains have been turbocharged this year by the public float of SpaceX, in which Google-parent Alphabet and Nvidia both own stakes. The rocket company absorbed Elon Musk’s xAI before going public, making the promise of AI data centres in orbit a key part of its pitch. The listing delivered a huge uplift for existing investors. The ‘other income’ contribution to profits last quarter was more than double the roughly US$69 billion boost in the previous three-month period. The expected public offerings of Anthropic and OpenAI would see the phenomenon continue in the coming year. In the most recent round of earnings reports, pre-tax profits reached a record at several of the Big Tech “hyperscalers”. But the main source of these increases at Alphabet and Amazon was ‘other income’ from gains on SpaceX and Anthropic, rather than new business lines or fresh cash generation. The companies only revalue their holdings in private companies following a fresh funding round, while publicly traded holdings are adjusted each quarter. This can create a distorted impression of earnings growth for investors and analysts looking at high-level financial metrics that do not adjust for these factors company by company. https://tinyurl.com/mvdet4fw

Snowflake raises annual sales forecast as AI coding adoption grows.

Shares of Snowflake rose more than 20% late Wednesday after the database provider surpassed its earlier forecast for product revenue—its most closely watched sales metric—by more than US$70 million for the second quarter ending July 31. Snowflake is benefiting as more of its customers use its AI products, which include an AI coding tool and a service for analyzing data to get answers to business questions. More than 9,100 customer accounts are using its AI coding tool, Cortex Code, up from 7,100 customer accounts last quarter. Snowflake’s overall product revenue grew 37% to US$1.49 billion in its July quarter compared to last year, up from 34% last quarter. The company raised its full-year product revenue target by US$230 million to US$6.07 billion, or 36% growth compared to last year, up from 31% last quarter. Snowflake expects product revenue to grow between 37% and 38% this quarter. One thing Snowflake hasn’t done lately is share specific figures about how much revenue its AI products are generating. The company last spring set an internal target of $100 million in annual recurring revenue—the value of its contracts over the next 12 months—and CEO Sridhar Ramaswamy said on an earnings call last December that it had reached this goal. https://tinyurl.com/482wy67b

Broadcom projects AI chip revenue to double in 2027 and in 2028.

Broadcom reported 86% higher revenue of US$29.6 billion in the quarter ending Aug. 2, a sign that its business helping design AI chips for companies such as Google, Meta and OpenAI is continuing to explode. On an earnings call Broadcom CEO Hock Tan projected that Broadcom would double AI revenue to US$115 billion in 2027, from an expected US$57 billion this year. And Broadcom has a “line of sight” suggesting AI chip revenue could double again to $230 billion in 2028. Broadcom was on track to exceed US$30 in earnings per share in 2028. Last year, Broadcom’s earnings per share was just below US$5 a share. In the most recent quarter, Broadcom’s AI chip revenue jumped 221% to US$16.7 billion. Broadcom projected that growth in AI chip revenue would accelerate 236% in the quarter through October. https://tinyurl.com/3av7ey9s

HPE raises forecasts on AI demand, shares fall on supply concerns.

Hewlett Packard Enterprise raised its annual forecasts on Wednesday after reporting quarterly results that beat Wall Street expectations, as surging AI-related demand across networking and servers drove growth ‌across its business. However, the company’s shares slipped more than 3% in extended trading, weighed down by persistent supply constraints. In an interview with Reuters, CFO Marie Myers noted that supply remained constrained, with memory being the main bottleneck, ⁠followed by NAND, CPUs and drives, though HPE has signed longer-term supply agreements to improve access to components. “Demand is far outstripping supply.” HPE’s inventory of US$11.82 billion at the end of July was up from US$7.16 billion last year. The increase reflects “higher commodity costs and targeted purchases to support increased orders and increased backlog,” Myers said. “It’s clear to us that the market, the tailwinds in AI, particularly the adoption of AI in the enterprise, is really starting to settle in,” Myers said. HPE expects enterprise adoption of AI to support growth beyond ‌the current ⁠fiscal year. It also announced an expansion of its collaboration with Oracle to help scale the software giant’s global AI infrastructure, deploying HPE Juniper Networking equipment across Oracle’s AI data centers. The AI server maker has also issued Oracle warrants to purchase shares of HPE common stock. https://tinyurl.com/y4pbyrsd

Dell raises sales forecast on AI server demand.

Dell said Tuesday that revenue for its quarter ending in July rose 58% from the year-ago quarter to US$47 billion, beating the high end of its own forecast. The maker of personal computers and servers for AI data centers increased its outlook for the year ending in January 2027 by US$25 billion to US$192 billion, which would represent a 70% increase from the year-ago. Dell attributed the rosier forecast to sales of its AI servers. The company anticipated US$74 billion from AI server revenue in its current fiscal year, up 200% year on year. Its server customers include the Pentagon, CoreWeave, and Nscale. Net income for the quarter rose 255% to US$4.1 billion. Dell’s stock rose 8% during extended trading hours Tuesday. It has risen 237% this year. The report follows another earnings windfall from Nvidia, the main beneficiary of hardware tied to AI. https://tinyurl.com/yrk63f5a

Palo Alto Networks revenue growth accelerates as companies buy more security for AI.

Palo Alto Networks lifted revenue 34% to US$3.4 billion in the three months ending in July. CEO Nikesh Arora attributed the growth, which accelerated from 31% the quarter prior, to growing demand for cybersecurity from companies seeking to defend against new AI-powered threats. Still, Palo Alto Networks reported a net loss of US$282 million in the quarter, compared with US$254 net income a year prior, stemming in part from costs associated with its US$25 billion acquisition of CyberArk earlier this year. The company touted growing sales of its AI-powered software: Arora said that one product released last spring that focused on securing companies’ AI agents has grown to over US$120 million in annual recurring revenue. Arora also said that the company’s XSIAM product, which uses AI to detect and flag potential security threats, surpassed US$700 million in annual recurring revenue, up from US$400 million a year earlier. Arora also announced that Palo Alto Networks on Tuesday closed an acquisition of Console, a startup that sells software that handles employees’ IT support requests directly in Slack using AI agents. The deal reflects Arora’s growing appetite for deals that expand Palo Alto Networks beyond its core business of cybersecurity into adjacent IT management categories; The Information reported last month that Arora was seeking out such deals. “AI is underscoring the urgent need for unified platforms,” Arora told shareholders, pointing to OpenAI’s recent AI security breaches as evidence of the need for more protection. “If [companies] deploy agents and those agents go rogue, how do we make sure our agents don’t run to Hugging Face?” https://tinyurl.com/3ed2h2ju

OpenAI starts letting some customers pay only when the AI works.

OpenAI in recent months has started giving some major customers the option of paying only when its AI completes tasks like handling customer support interactions. The move comes as Salesforce and a host of other AI providers are beginning to try to charge customers only when the AI works, otherwise known as outcome-based pricing, to help customers justify ever-larger purchases. Perhaps anticipating future disputes about the new pricing model, payment provider Stripe has published best practices for how to pursue such a model and how to avoid misunderstandings, the report said. https://tinyurl.com/5fa2s2vt

Microsoft discloses Azure revenue in accounting overhaul.

Microsoft’s cloud unit Azure’s revenue was US$101.9 billion in the twelve months ending in June, Microsoft disclosed Wednesday, up 40% from the year prior. Microsoft made the disclosure, and revealed it will start reporting Azure’s revenue quarterly, as part of a broader overhaul of how it reports the financial breakdown of its various businesses. And it comes as Azure’s progress has become a key metric that investors watch to track Microsoft’s AI progress. Microsoft previously included Azure’s results in a broader business segment that included both cloud server rentals and software subscription sales. While it reported Azure’s percentage growth, it didn’t break out the revenue regularly. The new accounting structure will make Microsoft’s Azure disclosures more directly comparable to similar revenue reported by rival cloud providers Amazon and Google. Like those company’s disclosures, Microsoft’s new Azure revenue number will reflect how much customers are spending on cloud infrastructure. Microsoft will also start disclosing the total quarterly revenue of its “M365 Cloud” products, which include its flagship Office 365 software as well as its sales AI products including its suite of Copilot AI tools. CEO Satya Nadella said Wednesday the overhaul was an effort to give investors more transparency and demonstrate how new AI products are changing its business. Nadella previously told shareholders in July that Azure revenue surpassed US$100 billion in the year that ended in June. More broadly, Microsoft will now report earnings across two segments: one called Agents and Infra, which will include its enterprise software and its Azure cloud computing businesses; and another called Devices and Consumer, which will cover its Xbox gaming unit, Surface hardware sales, and consumer Windows software, as well as advertising revenue from its websites including Bing and LinkedIn. https://tinyurl.com/4jez29mb

Uber to cut 10% of employees.

Uber is laying off 10% of its workforce, or about 3,400 people, in an effort to streamline the organization by reducing management layers. In making the announcement, Uber CEO Dara Khosrowshahi told staffers that “the vast majority” of remote employees needed to move to an office, reducing the proportion of people who can work fully remotely to about 1%. Khosrowshahi cited internal surveys showing that employees feel “too much work requires coordination across teams, debates take too long and decision-making rights are unclear.” As a result, the number of “micro-teams” of only one or two people were being cut in half. He didn’t make any reference to AI as a factor. Uber’s workforce has grown to 34,000 as at Dec. 31, from 22,800 at the end of 2020. Over the same period, Uber’s revenue has jumped from US$11 billion to US$52 billion. https://tinyurl.com/bdcse8fh

The Dream of Quantum Computing is both dead and alive.

People building quantum computers promise the machines will, very soon, do extraordinary things. Although it won’t make sense to use them to send an email or run spreadsheets, they’ll have the potential to help find cures for diseases, curb climate change and revolutionize finance. Many technologists also say that once a quantum computer gets powerful enough to crack the complex math behind encryption protocols, it will wreak all manner of cybersecurity havoc. At that point, if proper security measures aren’t in place, the digital gates that protect banks, government secrets, Bitcoin wallets and everything else online will be worthless. A recent paper from Google researchers predicted that this cybersecurity apocalypse, known as Q-Day, could come as early as 2029. Similar to AI researchers, the physicists pursuing quantum computers say the safest way to develop such a potentially dangerous technology is to have them do it before someone else does. Still, there are lemming tendencies. If one quantum company finds success with a technical approach or marketing tactic, others follow suit. Case in point: IQM introduced its chandeliers after IBM Corp. and Google started using their own. Another trend in the industry has been an interest in special purpose acquisition companies, financial instruments that allow companies to list with less scrutiny than conventional initial public offerings. SPACs became a market craze during the pandemic and were particularly popular among businesses working in speculative areas such as crypto or electric-scooter sharing. (Companies that used them were also notorious for crashing and burning.) IQM is one of several makers of quantum computers to go public via a SPAC, raising $1.8 billion in its listing in early July. Other quantum-related listings have led to Wall Street tickers with comic book names including Xanadu Quantum, Infleqtion and Quantinuum. Quantum computing has also attracted vast amounts of funding from national governments, which have committed about US$50 billion to the field this year, according to Quantum Business Network UG, an industry consultant. China represents almost a third of that. On May 21 the Trump administration announced plans to invest US$2 billion in quantum companies, with half going to a new IBM plant to make chips for the machines. Europe, which hasn’t been a serious competitor in tech since Nokia Oyj was one of the world’s most valuable companies back in 2000, is desperate not to miss this opportunity. The Finnish state funded IQM; German agencies bought multiple quantum computers; the Danes acquired a massive machine named after a Norse god. Here quantum computing is less an investment thesis than a matter of geopolitical game theory. Europe has watched ties across the Atlantic fray with distressing speed. If quantum computers do work, Europeans don’t want to be on the outside looking in or in a position where the US, a longtime reliable ally, might not share access. https://tinyurl.com/yc7zdbbm

OpenAI releases GPT-6 Astra model, suggests it could be AGI.

OpenAI on Thursday started releasing its new GPT-6 Astra model, billing it as a major advance in commercial tasks from financial modeling to engineering design. The new flagship model, which OpenAI said also has significant cybersecurity capabilities, will initially be available to a limited number of organizations, including those that are part of OpenAI’s Daybreak Access cybersecurity program, the company said in a blog post. Astra will be made available in the coming days to ChatGPT Plus, Pro, Business and Enterprise users, as well as through the OpenAI application programming interface and Amazon Web Services, the company said. OpenAI said that the new model showed significant improvements in areas like using a computer, software programming and professional work like creating presentations and spreadsheets. The company added that the model performs at the highest level on benchmarks including FrontierMath Tier 1, which tests mathematics ability, ARC-AGI 3 for reasoning ability and ExploitBench for cybersecurity. In a press briefing, OpenAI cofounder and President Greg Brockman said that the question of when the industry achieves artificial general intelligence, or AI that’s as good as humans at most tasks, has been a fuzzy one. However, if we “fast forward a couple years and look back,” he said, the milestone of achieving AGI “might be around this time and about this model.” OpenAI said in its blog post Thursday that its “evaluations found Astra’s written reasoning harder to monitor than GPT-5.6 Sol’s.” OpenAI said that it attributed this to “Astra’s greater control over written reasoning on simpler tasks and ability to solve problems with fewer written steps.” However, it added that GPT-6 Astra wasn’t as good at hiding its thinking process on more complex tasks. GPT-6 Astra will be 2.5 times more expensive than OpenAI’s previous flagship model, GPT-5.6 Sol, at US$10 per million input tokens and US$50 per million output tokens. It’s also the same price as Anthropic’s recently-released Fable 5.1 model, showing how Anthropic is still setting prices for the rest of the industry. However, Brockman hinted that OpenAI and the industry as a whole could eventually move away from per-token pricing for AI models. “Pricing tokens doesn’t make any sense,” he said. “Our tokens are not the same as our competitors’ tokens, they’re not the same between different model families.” He added, “What you actually want is…the price per task is what matters.” https://tinyurl.com/y4dr9htp

Anthropic touts better performance, lower cost in upgraded top models.

Anthropic released upgrades to its top AI models that it said deliver better performance and lower cost. Claude Fable 5.1 will cost an estimated 25% less than Fable 5 for token-based billing on what Anthropic described as typical workloads, the company said in a blogpost Tuesday, adding that the savings could be much larger for some highly agentic work. Fable 5.1 will also eventually include a new data system called Enterprise Frontier Safeguards that Anthropic said will give customers the same privacy as a zero data retention policy while preventing adversarial use. That will become available to business customers beginning later this fall, it said. Anthropic also released Claude Mythos 5.1, which is the same underlying model designed for more advanced cybersecurity and life sciences work, through its trusted access programs. The company said it also has improved its cybersecurity safeguards to reduce the number of instances where it flags benign content as dangerous. https://tinyurl.com/4cj5dnpk

OpenAI plans to limit Astra’s cybersecurity capabilities.

OpenAI plans to limit access to the most advanced cybersecurity capabilities of its forthcoming Astra model, according to a Tuesday blog post. The company said that it believes Astra meets the Critical cybersecurity threshold under its Preparedness Framework, meaning that the model can find previously unknown security flaws and ways to exploit those flaws across many well-protected systems without a person guiding each step. It is the first model the company is designating at this level, OpenAI said. Astra’s advanced cybersecurity capabilities will initially only be available to a smaller group of testers. The company plans to later expand access to companies through its Daybreak Blue program. The announcement comes after OpenAI paused work on Astra last month due to safety concerns. https://tinyurl.com/y5f687c8

Meta unveils new AI model in accelerated effort to catch rivals.

Meta Platforms announced Wednesday that it’s rolling out another AI model, Muse Spark 1.3. Meta’s chief AI officer, Alexandr Wang, said in an X post that Muse Spark 1.3, which is available through its coding product Muse Code and the Meta model API, is its “most capable model yet,” with improved agentic and coding capabilities. The model uses about 25% fewer tokens than its predecessor, Wang wrote. While Meta is charging the same per token price as Muse Spark 1.2, the model’s lower token usage makes it cheaper to run for the same amount of work. The release reflects Meta’s accelerated efforts to catch up with models and coding tools from rivals Anthropic and OpenAI. Meta launched the first Muse Spark model in April, then released Muse Spark 1.1 in early July and the 1.2 version less than a month ago. In a blog post, Meta touted the new model’s performance on several key benchmarks. On knowledge work, professional tool use and agentic computer use, Muse Spark 1.3 ranked slightly higher than OpenAI’s GPT-5.6 Sol (max). On the DeepSWE benchmark for long-horizon agentic coding, Muse Spark 1.3 outperformed both OpenAI’s GPT-5.6 Sol (max) and Anthropic’s Opus 5 (max). “This is the biggest jump we’ve made so far on coding and agentic work,” CEO Mark Zuckerberg wrote in an X post. He also teased Meta will release an AI model, internally called Watermelon, next. The Information reported last week that Meta is targeting October for Watermelon’s release. https://tinyurl.com/38wjvmxk

Tesla debuts Cybercab at launch event.

Tesla on Thursday held a launch event in Austin, Texas, for its Cybercab, the company’s first vehicle designed to operate without a steering wheel or brake pedals. The company invited a select number of Tesla fans to the event. Sawyer Merritt, a major Tesla influencer, shared a video from the event on X, saying there were about 400 people. The company updated its Robotaxi app on Thursday morning by adding the Cybercab to its terms of service and previewed a feature on the App Store designed to allow passengers to choose between selecting a Model Y or Cybercab for a Robotaxi ride. On Monday, Tesla registered 45 Cybercabs with the Texas Department of Motor Vehicles – a necessary step before the company can begin using the cars in its commercial fleet. In total, the company has 420 vehicles registered for its Robotaxi hubs in Texas, which include Austin, Dallas, and Houston. https://tinyurl.com/48r3kjhd

OpenAI’s advertising business hits US$1 billion annualized revenue run rate.

OpenAI’s advertising business has reached US$1 billion in annualized revenue run rate, the company said Monday, suggesting it is likely to fall well short of the US$2.4 billion in ad revenue it had projected for 2026. The ChatGPT creator launched ad sales in the chatbot in February, as a way to generate money from the hundreds of millions of people who use ChatGPT without pay. Earlier this year, OpenAI bet ads would become the largest driver of its revenue by the end of the decade. Lately, however, OpenAI has focused heavily on building a business selling to enterprises. In a blog post on Monday, OpenAI said advertising was “one pillar of OpenAI’s diversified business model, alongside consumer subscriptions, enterprise offerings and usage-based APIs,” the term by which developers can plug into OpenAI’s AI models. https://tinyurl.com/3kk455rd

FTC sues Amazon, alleging ad price manipulation.

The Federal Trade Commission and attorneys general from 22 states sued Amazon for overcharging its advertising customers by allegedly manipulating ad auctions. Advertisers promoting their products on Amazon’s marketplace buy ad spots where the price is set in auctions. In the complaint, the FTC and the states allege Amazon added an extra charge to bids in order to raise the price that advertisers paid above what one employee said could be achieved by “advertiser competition” alone. Amazon said in response that the FTC’s claim “fundamentally misunderstands how advertisers operate,” noting that “advertisers adjust bids based on real-world performance.” It said “in no scenario does an advertiser pay more than their bid.” Amazon also said it was “patently false” that Amazon tried to deceive anyone. The FTC also presented evidence that Amazon took several steps to conceal that it was inflating auction prices from advertisers. The complaint says the practice started in 2018 and continues on to this day, and that it likely resulted in advertisers overpaying by tens of billions. The lawsuit is reminiscent of antitrust cases against Google which accused that company of also manipulating ad auctions, in both its search advertising business and its advertising technology business. Google lost both cases. The suit is at least the third lawsuit the FTC has filed against Amazon in the past few years, including a major case alleging the tech giant illegally monopolized ecommerce. As part of that case, the FTC called the advertising fees Amazon charges merchants “monopoly rents.” That case is expected to go to trial next year. Amazon settled a separate case with the FTC last year; this one centered on Amazon’s alleged practice of tricking customers into signing up for Prime accounts and making it difficult to cancel. https://tinyurl.com/4js8ntyp

Google avoids break-up in Ad-Tech Anti-Trust case.

A federal judge ruled that Google does not need to divest parts of its ad-tech business, despite her earlier ruling that the company monopolized parts of that market, the second time it has avoided an antitrust-inspired breakup. In April 2025, US District Court judge Leonie Brinkema found Google guilty of monopolizing parts of the market for advertising technology, through which Google helps independent websites sell advertisements. The US government argued Google should be forced to sell parts of its ad tech business but Brinkema instead ordered “behavioral remedies,” forcing Google to change how it operates rather than divesting the businesses. It’s not clear exactly what behavioral remedies she’s referring to, as her full opinion won’t be public until later this month, but the government had suggested making Google’s technology more interoperable with products from competitors. Google, for its part, has suggested a similar solution in its proposed remedies, though the two parties differ in some details on how this might be implemented. Google last year avoided a breakup in a separate antitrust case brought by the government relating to its search business. “We’re very pleased the Court rejected the DOJ’s proposal to break apart tools that help small businesses reach new customers and grow,” said Lee-Anne Mulholland, vice president of regulatory affairs at Google, in a statement. A spokesperson for the Department of Justice said its antitrust division is “pleased that the court ordered substantial relief in the Google Ad Tech case” and that it was “evaluating appropriate next steps.” “We are one step closer to restoring competition and bringing relief for the American people in online advertising markets,” the spokesperson said. https://tinyurl.com/9sevtnp5