Daily News Coverage: Chip Stocks Drop Amid Inflation Fears; OpenAI Inks 12GW NVIDIA Deal
The CODEW Daily News Coverage | August 19, 2026
Good morning, people! Chip stocks led a broad market decline Tuesday as rising long-term yields and inflation fears weighed on the AI trade, while OpenAI committed to 12 gigawatts of NVIDIA infrastructure by 2030. Alibaba open-sourced Qwen3.8-27B, AMD is reportedly in talks with Google for next-generation TPU development, and a critical NASA spacecraft command console flaw underscores the cybersecurity risks in critical infrastructure. Here's today's briefing from The CODEW's newsroom.
THE LEAD
Chip Stocks Tumble as Rising Yields and Inflation Fears Hit AI Trade
What happened: U.S. semiconductor stocks led a broad market decline Tuesday as rising long-term yields and inflation concerns weighed on the AI trade. The S&P 500 fell 0.7%, the Nasdaq dropped 1.7%, and the Dow declined 116 points. NVIDIA fell 2.3% to $219.74, AMD dropped 4.3%, Broadcom declined 3.2%, Micron tumbled 7%, and Intel lost 6.6%. SanDisk plunged 9%. Long-term yields, though off recent highs, remain near multi-year peaks as inflation risks and a surge in AI company bond issuance have heightened uncertainty around term premiums.
Why it matters: The semiconductor selloff reflects growing investor anxiety about the sustainability of AI infrastructure spending amid rising borrowing costs. Memory chip stocks were particularly hard hit — a sign that the market is repricing expectations for AI hardware demand. The Magnificent Seven also declined broadly, with Meta down 4.4% and Tesla off 0.7%.
Who is affected: AI chipmakers (NVIDIA, AMD, Intel, Broadcom, Micron), hyperscalers, and enterprise AI buyers facing rising infrastructure costs.
What to watch next: Whether the chip selloff deepens or stabilizes as investors assess the durability of AI infrastructure spending against rising capital costs.
AI & Infrastructure
OpenAI Commits to 12GW of NVIDIA Infrastructure by 2030
OpenAI committed to deploying 12 gigawatts of NVIDIA AI infrastructure by 2030, according to a report. Existing and planned commitments represent approximately 12 GW of NVIDIA compute, with the opportunity to expand to approximately 16 GW if NVIDIA extends the PORTS Pike agreement beyond the initial 4.25 GW. At these levels, the opportunity represents approximately $600 billion in NVIDIA compute by 2030.
The commitment comes as NVIDIA, OpenAI, and SoftBank collaborate on converting Ohio's abandoned uranium enrichment plant into what could become the world's largest AI super-factory, with ultimate power planning of up to 8 GW, enough to power approximately 1.5 million GPUs simultaneously. NVIDIA has invested $1.5 billion in SoftBank's subsidiary and is providing a 20-year credit guarantee, with OpenAI signing a 20-year lease to fully utilize the compute capacity.
Why it matters: Huang Renxun has noted that the next bottleneck for AI is not chips but land, power, and factory space. This move secures scarce resources while ensuring the facilities can only use NVIDIA chips. OpenAI alone is expected to generate $600 billion in revenue by 2030.
Alibaba Open-Sources Qwen3.8-27B, Bringing Frontier AI to Consumer GPUs
Alibaba has released the open weights of Qwen3.8-27B, a 27-billion-parameter model that runs on a high-end PC. The release comes as part of the broader Qwen3.8-Max announcement, which features a 2.4-trillion-parameter MoE architecture. The model is already available on Hugging Face.
Why it matters: Qwen3.8-27B brings frontier-level AI capabilities to consumer-grade hardware, accelerating the democratization of AI development. The release adds to the growing portfolio of open-weight models from Chinese labs that are matching or exceeding U.S. frontier models at a fraction of the cost.
Semiconductors
AMD in Talks with Google for Next-Generation TPU Development
AMD is reportedly in talks with Google to develop next-generation Tensor Processing Units (TPUs). If confirmed, this would mark AMD's first significant entry into the AI ASIC space. AMD's advanced packaging, SoIC, and CPU core integration capabilities are considered key factors in the selection. Google has historically worked with Broadcom on TPUs; the shift suggests Broadcom may be unable to provide the specific capabilities Google is seeking for its next-generation chip.
The potential partnership comes as TSMC's advanced packaging capacity remains constrained due to surging AI chip demand, prompting AMD to expand its packaging partnerships with Taiwanese suppliers including ASE Group and Powertech Technology. AMD has committed to investing more than $10 billion in Taiwan's semiconductor ecosystem.
Why it matters: A Google-AMD TPU partnership would be a major validation of AMD's AI capabilities and a significant competitive threat to Broadcom's dominant position in AI ASICs. It would also give Google an alternative to NVIDIA for AI training workloads.
TSMC CoWoS Orders Overflow as Intel Steps In to Support Common Customer
TSMC's CoWoS advanced packaging capacity is fully booked amid surging AI chip demand, with reports that some backend orders have overflowed to Intel's Malaysia facility. The arrangement breaks with traditional ecosystem conventions, as Intel and TSMC are normally direct competitors.
Why it matters: The CoWoS capacity crunch reflects the extraordinary demand for AI chip packaging, which has become a critical bottleneck in the AI supply chain. Intel's willingness to support a shared customer — reportedly a major AI chipmaker — underscores the severity of the capacity constraints and the industry's need for expanded advanced packaging capacity across multiple suppliers.
Chip Designer Velaura AI Valued at Over $1 Billion in Series A
Velaura AI, a chip design startup developing low-power technologies for AI data centers and physical AI applications, raised $110 million in a Series A round that valued the company at more than $1 billion. The round was led by Seligman Ventures with participation from Samsung Catalyst Fund, StepStone Group, and Maverick Silicon.
The startup's Titan Core platform targets greater efficiency and power savings in data center workloads. Velaura charges an upfront fee plus a royalty tied to power savings achieved — a structure similar to Arm's per-chip licensing model. The company is engaged with three of the four largest cloud providers.
Why it matters: Velaura's $1 billion+ valuation reflects investor conviction that power efficiency — not just raw compute — is becoming the defining competitive advantage in AI infrastructure. The Arm-style licensing model could create durable recurring revenue.
Enterprise Software
Salesforce Launches MuleSoft Agent Fabric and Omni Gateway to Tame Agent Sprawl
Salesforce released new features for MuleSoft Agent Fabric and a new solution called MuleSoft Omni Gateway, designed to help enterprises manage the proliferation of AI agents across multi-platform environments. The tools address two emerging challenges: "AI agent sprawl" and "AI cost management".
Salesforce noted that token consumption is expected to increase 24x over four years. Uber reportedly used its entire 2026 AI budget in just four months, while AT&T's internal AI consumption surged from 1 billion to 27 billion tokens per day over 18 months without corresponding business value.
The new features include Agent Scanner for discovery, orchestration capabilities, governance controls, and monitoring across discovery, orchestration, governance, and monitoring layers.
Why it matters: As enterprises move from AI agent pilots to production deployments, management and cost control are becoming critical challenges. Salesforce is positioning MuleSoft as the control plane for multi-agent enterprise environments.
Cybersecurity
NASA JPL Spacecraft Command Console Flaw Allowed Unauthenticated Commands
A critical vulnerability (CVSS 9.4) was discovered in NASA Jet Propulsion Laboratory's open-source AIT-GUI software, the browser-based console used to command spacecraft and scientific instruments. The flaw, tracked as GHSA-p9r8-2q67-fp86, allowed unauthenticated remote attackers to issue arbitrary commands to spacecraft systems. The issue has been fixed in version 2.5.2.
The vulnerability combined multiple weaknesses: no login requirement, no session check, no CSRF protection, and a server hardcoded to listen on all network interfaces (0.0.0.0) rather than localhost. Even behind a firewall, an operator visiting a malicious web page could have sent commands to the interface.
"A web GUI used to drive spacecraft and instrument commanding shipped a server that listens on every network interface, asks nobody for a password, and can be steered by any web page an operator happens to open," said Cycode researcher Yuval Elbar.
Why it matters: The vulnerability highlights the security risks in operational technology (OT) and critical infrastructure — where web interfaces control physical systems. The fact that spacecraft command systems lacked basic authentication underscores the gap between enterprise security standards and critical infrastructure security.
Critical GitLab Zero-Click Flaw Allows Unauthenticated Project Deletion
A critical zero-click vulnerability (CVE-2026-19478, CVSS 9.4) in GitLab Community and Enterprise Editions could allow remote attackers with no account or credentials to manipulate or delete publicly accessible projects and user data via GraphQL. The flaw affects all versions from 18.2 before 18.11.11, 19.0 before 19.0.8, 19.1 before 19.1.6, and 19.2 before 19.2.4.
GitLab has withheld technical details for 90 days after the patch release, making it difficult for organizations to detect exploitation. A second CSRF flaw (CVE-2026-19650, CVSS 7.1) affects the same versions.
Why it matters: GitLab is widely used for source code management and CI/CD pipelines across enterprise development organizations. A zero-click vulnerability that enables unauthenticated project deletion represents a critical supply chain risk — attackers could potentially delete source code repositories, disrupt development operations, or inject malicious code.
What to Watch
- AMD-Google TPU partnership confirmation: If confirmed, this would be AMD's first major AI ASIC win and a significant competitive shift in the AI chip market.
- AI infrastructure spending sustainability: The chip selloff reflects growing concerns about AI capex sustainability amid rising borrowing costs — watch for hyperscaler commentary on infrastructure spending plans.
- OpenAI NVIDIA infrastructure deployment: The 12 GW commitment by 2030 represents approximately $600 billion in NVIDIA compute — watch for execution and financing details.
- GitLab vulnerability exploitation: With technical details withheld, watch for signs of active exploitation against self-managed GitLab instances.
- Korean AI model competition: LG's 750B-parameter K-ExaOne 2.0, SK Telecom's A.X K2, and Upstage's Solar Open 2 are competing in Korea's "Independent AI Foundation Model" project.
Source Attribution
- Reuters — Chip designer Velaura AI valued at more than $1 billion in funding round (August 19, 2026)
- SecurityBrief — NASA lab console flaw allows unauthorised commands (August 19, 2026)
- Dark Reading — Critical GitLab Zero-Click Flaw Poses Mitigation Challenges (August 19, 2026)
Reviewed by Erwin Castro
on
Wednesday, August 19, 2026
Rating: