US national debt has reached $40 trillion and now exceeds 123% of gross domestic product, marking a new benchmark for federal borrowing relative to the size of the economy.

Polymarket's U.S.-regulated prediction market has climbed to $111.4 million in open interest, nearly doubling since the 2026 FIFA World Cup ended in mid-July. The product now represents 54% of total volume on Polymarket, a notable shift for a platform that only launched after Polymarket's December 2025 acquisition of QCEX. The rise stands out against broader industry cooling. Since the World Cup concluded on July 19, combined open interest across major prediction venues, including Kalshi, has fallen by roughly 20%. Polymarket US moved in the opposite direction, extending gains through August and reaching $111.4 million by Aug. 22. The World Cup itself drove outsized activity across prediction markets. Polymarket's soccer category generated more than $2 billion in volume during the tournament's first 10 days. Kalshi's peak combined open interest during the event was around $2 billion before sliding to roughly $1.5 billion by late July. Polymarket US differs structurally from the original Polymarket. While the core platform is crypto-native and uses on-chain collateral, the U.S. version operates exclusively with fiat USD contracts under full KYC compliance. It does not rely on crypto collateral or anonymous wallets. Its growing share of Polymarket volume signals demand from users comfortable with identity checks and dollar-denominated exposure, rather than short-term, high-churn speculation. Positioning has also shifted as the tournament cycle ended. World Cup match-by-match markets tended to open and close within hours, reflecting short-dated contracts tied to specific games. Post-tournament, remaining contracts are generally longer-dated, a dynamic that can lift open interest even as trading volume moderates. Kalshi still leads by a wide margin in absolute open interest—about $1.5 billion versus $111.4 million for Polymarket US—but the trajectories have diverged, with Kalshi trending lower while Polymarket US continues to build.

ME News reported that on Aug. 26 (UTC+8), the U.S. Navy cleared the primary shipping corridor in the Strait of Hormuz—the Traffic Separation Scheme (TSS) between Iran and Oman—enabling higher tanker traffic. U.S. officials said that over the past several months, Navy underwater drones conducted systematic sweeps of the area, flagging more than 100 suspected mine-like objects. The items were later removed or destroyed through joint efforts by the U.S. military and private companies. In the past 30 days, more than 500 vessels transited the strait's southern lane, with about 2% reportedly targeted by Iranian drones or missiles. Trump said any vessel attempting to lay mines would be "immediately and systematically destroyed." U.S. officials added that with the main lane reopened, ships can now move through the Strait of Hormuz in both directions under U.S. protection, supporting larger volumes of oil flows to global markets. (Source: BlockBeats)

OpenAI has introduced a new in-house AI chip called Jalapeño, positioning it as a move to curb reliance on third-party hardware suppliers and improve efficiency for AI workloads, according to Blockchain.News. Key claims around Jalapeño differ slightly by source. CryptoBriefing cited Broadcom's CEO saying the chip matches Nvidia's Blackwell architecture while delivering roughly a 50% cost advantage. Broadcom has reportedly worked with OpenAI on custom silicon, giving its leadership visibility into the design and manufacturing economics. Cryptopolitan described a separate emphasis from OpenAI, reporting the company claims Jalapeño outperforms Nvidia's GB300 on inference workloads. Inference refers to running a trained model to generate outputs, distinct from training. The distinction is significant because inference costs typically scale with user demand, making gains in inference efficiency directly relevant to operating expenses at large deployment volumes. The difference between "matches" and "outperforms" reflects sourcing and framing more than a direct conflict: Broadcom's comparison focuses on cost relative to Blackwell, while OpenAI's claim highlights inference performance versus GB300. For buyers allocating compute budgets, both cost-per-performance and inference throughput matter, even though they measure different dimensions. OpenAI's push into custom hardware follows a broader industry trend. Google, Amazon, and Meta have all developed proprietary chips to reduce dependence on Nvidia GPUs, which remain the dominant standard for AI training and inference. With demand for AI compute fueling heavy data-center investment over the past two years, a credible alternative that lowers inference costs could influence infrastructure planning—especially for companies operating at OpenAI's scale. Market impact: A competitive OpenAI-designed chip could challenge Nvidia's pricing power if the cost and performance assertions hold up under independent benchmarks. Broadcom may also draw increased investor attention as a reported partner in the custom-silicon effort. Lower-cost inference could also reduce operating expenses for firms building AI-driven products across technology and crypto-adjacent sectors, including trading systems and blockchain analytics that rely on machine learning. Jalapeño becomes the latest entrant in the race to develop alternatives to Nvidia's AI hardware. Broader adoption will likely depend on the availability of third-party performance data and real-world deployment results. FAQ - What is the Jalapeño chip? A custom AI chip developed by OpenAI for AI compute workloads, including inference. - How does it compare with Nvidia? Reports vary: Broadcom's CEO said it matches Blackwell at about half the cost; OpenAI has said it beats GB300 on inference. - What is inference? Running a trained model to produce outputs, separate from training, and often the main driver of ongoing operating costs. - Why build a custom chip? To reduce reliance on suppliers such as Nvidia and potentially lower the cost of operating large-scale AI services—an approach also pursued by Google and Amazon. Originally reported by AltcoinGordon; written by Daniel Foster. Republished with permission. View the original on AltcoinGordon.

Aug. 26 — Axios reported that the U.S. Navy has cleared the main shipping lane in the Strait of Hormuz, the Traffic Separation Scheme (TSS) between Iran and Oman, to enable more oil tankers to transit. U.S. officials said Navy underwater drones carried out systematic surveys over recent months and flagged more than 100 suspected mine-like objects. The U.S. military then worked with private companies to remove or detonate the devices. Over the past 30 days, more than 500 vessels have passed through the southern lane, with about 2% coming under attack from Iranian drones or missiles, according to the officials. Former President Donald Trump said any vessel that re-mines the area would be "immediately and systematically destroyed." U.S. officials added that with key routes resuming operations, ships can now move through the strait in both directions under U.S. protection, supporting higher oil flows into global markets.

Marvell has struck a custom-chip partnership with Google and will grant Google warrants for up to 58.97 million shares, with an exercise price of $206.58 per share. For fiscal 2027 first quarter, Marvell reported revenue of $2.42 billion, up 28% year over year. Data center revenue accounted for 76% of the total. The company forecast fiscal Q2 revenue of $2.7 billion, which would represent 34.6% year-over-year growth.

ETF inflows reached $2.26bn over six consecutive positive trading days.

AMD has crossed a notable threshold in the long-running x86 CPU battle, with client-market share now above 30%, according to Mercury Research. The figure marks a steep climb from 21.1% two years ago. Over the same period, Intel's share fell to 69.7%, its lowest level since 1995. Mercury Research president Dean McCarron said the data reflects "client share" only, covering desktop and mobile CPUs. The numbers also come with important caveats. The x86 share calculations exclude ARM-based processors, an increasingly meaningful segment led by Apple, Qualcomm, and MediaTek. Mercury's ARM-inclusive estimate for PC client CPUs in Q2 2026 puts ARM at 15.3%, a new high, up 0.9 percentage points. On the demand side, Mercury pointed to a softening desktop market. Higher PC prices and tight GPU supply are weighing on end demand for desktop systems, a datapoint likely to resonate with crypto- and hardware-focused readers given GPUs' importance in certain compute workloads. Even so, AMD's x86 gains are not the main story behind its stock performance. Investors have largely been rewarding the company's AI and datacenter trajectory, where accelerators and GPUs are the key value drivers. That disconnect showed up in recent trading. On Aug. 24, 2026, AMD shares closed down 3.49% (16.51 points), then rose about 2% (9.15 points) in after-hours trading. Wall Street's core thesis remains centered on AMD's AI expansion. In today's semiconductor cycle, AI chips—not PC CPUs—are doing most of the heavy lifting for growth expectations and near-term valuation. Bottom line: clearing 30% of the x86 client CPU market is a meaningful competitive milestone for AMD, but its market value is more tightly tied to AI and datacenter opportunity than to PC CPU share alone. The GPU supply constraints flagged by Mercury remain a key thread to watch, shaping both PC demand and broader compute availability.

Two voice AI players, SoundHound and Cerence, have just posted fresh earnings that highlight very different investment profiles. SoundHound reported second-quarter revenue of $61.9 million, up 45% year over year. The company also raised its full-year revenue outlook to $230 million–$260 million and said it carries no debt on its balance sheet. Cerence delivered 12% revenue growth in its fiscal third quarter to $69.6 million. Adjusted EBITDA rose 50%, while free cash flow came in at $19.6 million. The company is projecting 41% full-year revenue growth. The results underscore a tradeoff for investors: SoundHound is emphasizing faster top-line momentum, while Cerence is pairing growth with stronger profitability and cash generation.