Ben Thompson, founder of Stratechery, argued in a recent interview that Nvidia’s exceptional profitability may be less durable than it appears as the AI spending cycle enters a more contested phase. His view centers on two pressure points. First, he said Nvidia has supported newer cloud providers, or “Neoclouds,” through equity stakes and roughly 25% backstops tied to commitments to keep buying Nvidia compute through 2030. That may help sustain GPU shipments, but Thompson said the risk does not disappear; it shifts back onto Nvidia if compute demand weakens or those buyers cannot keep purchasing. In his framing, that amounts to a hidden reduction in profit and functions like an indirect price cut. Second, Thompson said Nvidia’s energy-efficiency edge matters most when power is scarce. He argued that unexpectedly resilient U.S. electricity supply over the past two years — including natural gas generation in West Texas, restarted nuclear plants, and grid-related deployments by Elon Musk — gives hyperscalers such as Amazon and Google more time to improve in-house chips like Trainium and TPU. That, in turn, could erode Nvidia’s technical moat. Even if the current AI boom ends in oversupply and a market correction, Thompson said the resulting buildout of power infrastructure may still become the most durable legacy of the cycle.

Nvidia is extending its AI playbook beyond semiconductors and into financing, using its cash flow and funding capacity to shape how large-scale AI infrastructure gets built. Over the past week, the company reached agreements with Goldman Sachs, Blackstone, BlackRock and Apollo to support a GPU financing effort that could reach as much as $500 billion. It then said it would provide up to $105 billion in support for OpenAI’s Ohio data center project. Chief Executive Officer Jensen Huang said many frontier AI labs are growing faster than their balance sheets and long-term credit capacity can support, creating the need for outside capital to fund infrastructure. Nvidia also plans to back related loans with support covering as much as 25% and has continued making equity investments across the AI supply chain. As Google’s TPU and AMD step up competition, Nvidia’s technology edge in chips is facing pressure, and the company appears to be using a combined chips-and-capital model to deepen control over the broader AI ecosystem.

Liquid cooling is moving from an optional feature to core infrastructure for high-end AI servers as chip power consumption climbs, according to TrendForce. The research firm said liquid-cooling penetration for AI chips is projected to rise from about 33% in 2025 to 53% in 2026, then approach 60% in 2027. It linked that shift to rapidly evolving processors from NVIDIA, AMD and Google, with single-chip thermal design power already exceeding 1 kW and full rack-scale systems reaching several hundred kilowatts. TrendForce said NVIDIA remains the clearest force behind broader adoption. Its Vera Rubin platform has adopted a fanless, full-liquid-cooling design that extends beyond GPUs and CPUs to include CX9 NICs, busbars, power boards and optical transceiver modules. Even with a possible delay to the Kyber NVL144 platform, the firm expects overall GPU rack shipments to grow more than 30% year over year, while NVIDIA GB/VR rack shipments could double in 2026. AMD is also expanding from standalone GPU products to full AI platform systems, with its Helios rack-scale solution set to become a focus from the second half of 2026 and larger-volume shipments expected in 2027. TrendForce added that more than 80% of Google’s AI servers already use liquid cooling. On the supply side, Taiwanese companies including AVC, Auras and Jentech have secured positions in cold-plate and heat-spreader segments.

Sandisk used its Aug. 13 investor day to lay out a long-range financial model that sits well above market expectations and puts its in-house High Bandwidth Flash, or HBF, at the center of the story. The company said its long-term sustainable model, based on average performance across FY2028 to FY2030, calls for revenue growth in the mid-to-high teens, a Non-GAAP gross margin of about 80%, a Non-GAAP operating margin of roughly 75%, adjusted free cash flow margin near 50%, and capital intensity in the mid-single digits as a percentage of revenue. Mizuho Securities kept its Outperform rating on Sandisk and set a $1,900 price target, arguing that the company’s framework, buyback capacity, and exposure to AI infrastructure support upside. Sandisk also detailed the positioning of HBF, saying the product can deliver comparable read bandwidth to HBM at about one-eighth the cost, while offering 8x to 16x more capacity in a similar package footprint. The company tied that pitch to memory-heavy AI inference workloads and said HBF is meant to work alongside HBM rather than replace it.

A report cited by MarsBit and originally published by ITjuzi says China’s dexterous hand segment has logged 74 financing events involving 47 companies from January to Aug. 3, 2026, with disclosed funding totaling about RMB 28.51 billion. The report divides the field into two groups: 15 humanoid robot makers that develop dexterous hands in-house, and 32 third-party suppliers that sell complete hands or core components such as tactile sensors, micro motors, and precision screws. The funding gap between the two camps is wide. The 15 robot body makers accounted for 23 deals and about RMB 18 billion to RMB 20 billion in disclosed funding during the first seven months of the year, while the 32 suppliers completed 51 deals worth about RMB 8.51 billion. The report argues that in-house hand development has become a standard requirement for humanoid robot OEMs, but the deeper component stack remains largely outsourced. It also says the supplier side is fragmenting into three distinct tracks: complete dexterous hand module vendors, tactile sensing companies, and core drivetrain and transmission component makers. While complete-hand vendors are facing crowded competition, companies focused on tactile sensing and specialized parts are described as having stronger technical moats and a longer runway for commercialization.

SemiAnalysis used a recent podcast appearance to lay out an unusually bullish case for SpaceX’s data center and AI infrastructure strategy. The firm’s analysts argued that the economics of frontier-model inference have already shifted far enough to justify aggressive capacity buildouts: on their numbers, API inference running on GB300 clusters can generate about $100 million in annual revenue per megawatt, with gross margins above 85%. In that framework, the real advantage is not owning chips in the abstract, but delivering usable power and compute faster than rivals can. That is why they view SpaceX’s offer as scarce “emergency megawatts” rather than ordinary long-term cloud capacity. The discussion also focused on pricing, customer incentives, and execution. SemiAnalysis said SpaceX can charge roughly $50 million per megawatt per year for quickly delivered capacity that comes with a 90-day cancellation option, and still leave customers with meaningful margin. Google, Microsoft, and Anthropic were identified as the most relevant demand sources, while Microsoft was described as the clearest fit because of its OpenAI exposure and its near-term capacity gap. On the build side, the analysts said warehouse conversions, mobile turbines, cross-state power strategies, and supplier financing make the supply problem manageable. Their biggest concern sat elsewhere: if autonomous agents become powerful enough to alarm the public and policymakers, demand could be constrained by political intervention rather than engineering limits.

Google’s Gemini roadmap has come under fresh scrutiny after a model the company previewed at Google I/O in May still has not arrived by mid-August. Gemini 3.5 Pro was shown on a keynote slide with the message that it was coming the following month, yet Google has only released newer Flash variants, including 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. Google’s public position is that Gemini 3.5 Pro remains in closed testing with limited partners and has no confirmed launch month, while Gemini 4 has already entered pre-training. SemiAnalysis, an independent research firm focused on semiconductors and AI, has taken a sharper view, arguing that Gemini 3.5 Pro has effectively been canceled. The report says the missing flagship model adds to broader concerns around Google’s AI execution, especially as public benchmarks place Gemini 3.6 Flash around eighth or ninth depending on the ranking system, and as rivals continue to push in reasoning, coding, and agent use cases. At the same time, Financial Times reported that Google co-founder Sergey Brin has re-entered high-level Gemini strategy discussions. His renewed involvement comes alongside major leadership changes at Google DeepMind, senior departures, and growing debate over whether Google is allocating enough TPU capacity to its own frontier model work while also supplying infrastructure to Anthropic.

Alphabet lost about $16 billion in market value on Wednesday after an unusually concentrated round of leadership and talent changes inside Google’s AI ranks. Jeff Dean and Sanjay Ghemawat, the only two engineers to ever hold Google’s L11 Senior Fellow title over a 27-year span, both resigned the same day. Hours later, Nobel laureate Demis Hassabis said he would step back from day-to-day management of Google’s AI lab, with DeepMind CTO Koray Kavukcuoglu taking over operations tied to Gemini 4. The article argues that the bigger question is not simply who left, but what Google actually lost. Dean, Ghemawat, Oriol Vinyals, and Quoc Le are now tied to Discovery Loop, a public-benefit company focused on automating the experimental loop in machine learning research. Yet Google is not just watching from the outside: it is a founding investor in Discovery Loop and its exclusive cloud provider. The same pattern shows up elsewhere, from Google’s roughly 14% stake in Anthropic to its $2.7 billion Character.AI licensing deal that brought Noam Shazeer back before he left again for OpenAI. Even as Google looks weaker in frontier model rankings, the piece says its distribution, cloud infrastructure, custom chips, and equity stakes still compound underneath the org chart drama.