Daily News Coverage: AI Models Get Cheaper as Nvidia Expands Infrastructure Race
Ten Fast Reads: What Changed Today Across AI, Semiconductors, Cybersecurity, Enterprise AI, and Capital
Good morning! Anthropic and OpenAI launch cheaper models days after calling for an AI slowdown. ShinyHunters claims an FBI breach while hijacking Clop's leak site. Nvidia expands its AI factory ecosystem with LG, VAST Data, and SB Energy, and backs a 300MW Armenia data center. Nasdaq hits a record high on AI rally as Micron and SanDisk surge. Here's today's briefing from The CODEW's The Newsroom.
Anthropic and OpenAI Launch Cheaper Models Days After Calling for AI Slowdown
OpenAI cuts GPT-6 API pricing by 50% and opens safety evaluations to third parties; Anthropic releases Claude Opus 5.5 at ~40% lower cost — a juxtaposition exposing the tension between safety rhetoric and commercial competition.
What happened: OpenAI introduced additional tiers to its GPT-6 family — GPT-6 Sol and GPT-6 Luna — with API pricing reduced by 50% compared to GPT-5.6 promotional pricing. Anthropic released Claude Opus 5.5, describing it as a more token-efficient version that costs approximately 40% less to run than its Opus 5 model. Both releases came after Anthropic CEO Dario Amodei published an essay urging the industry to slow frontier development, a position OpenAI CEO Sam Altman endorsed.
OpenAI also announced it will let third-party organizations evaluate AI safety risks at earlier stages of model development — during training, evaluation, and rollout — rather than only before release. The company outlined priorities for effective third-party evaluation, including "strong independent mechanisms, scientific rigor, robust safety practices, and clear accountability." Lama Ahmad, who leads OpenAI's external safety review partnerships, said the company previously brought in evaluators mainly before model release to assess capabilities.
Why it matters: The releases reveal a fundamental contradiction at the heart of the AI industry's current moment: the same labs calling for coordinated slowdowns are simultaneously competing aggressively on price and efficiency, driven by pressure from cheaper open-weight models. For enterprise buyers, this means the cost of frontier AI capabilities continues to fall even as safety concerns mount — a dynamic that could accelerate adoption while complicating governance.
What's next: Whether the pricing pressure triggers a broader price war among frontier labs, and how OpenAI's earlier third-party evaluation framework is received by regulators and independent safety organizations.
Sources: Sohu.
ShinyHunters Claims Breach of FBI, Alleges Theft of Thousands of Employee Records
The claim follows the group's hijacking of Clop's dark web infrastructure — the FBI has not commented on the alleged breach.
What happened: The hacking group ShinyHunters says it has breached the Federal Bureau of Investigation and claims to have stolen data on thousands of FBI employees, according to Reuters. The bureau did not respond to repeated requests for comment. The claim follows a pattern of aggressive targeting by ShinyHunters, which recently breached Clop's dark web infrastructure and is now extorting the ransomware group itself.
Why it matters: If verified, a breach of FBI employee data would represent a significant national security incident. Even if unverified, the claim demonstrates ShinyHunters' willingness to target the highest-profile organizations — and its growing notoriety in the cybercriminal ecosystem.
What's next: Whether the FBI confirms the breach and what data was actually compromised.
Source: Reuters.
Nvidia Expands AI Factory Ecosystem with LG, VAST Data, and SB Energy Investments
LG's cooling and battery storage systems certified under DSX Ready; VAST Data partnership enables confidential computing; SB Energy investment reaches $3 billion.
What happened: LG Electronics' 2.5-megawatt coolant distribution unit and LG Energy Solution's battery energy storage system were certified under Nvidia's DSX Ready program, which validates components for next-generation AI data centers. Nvidia also announced a partnership with VAST Data to enable enterprises to deploy AI on their own sensitive data using Nvidia Confidential Computing, allowing businesses to run AI on proprietary data while protecting commercial secrets and models. Separately, Nvidia is investing an additional $1.5 billion in SB Energy, a SoftBank-backed data center infrastructure provider, bringing its cumulative investment to $3 billion ahead of SB Energy's IPO.
Why it matters: The announcements collectively illustrate how Nvidia is positioning itself as the central platform for AI infrastructure — not just supplying chips but orchestrating an ecosystem of cooling, power, data management, and confidential computing partners. The VAST Data partnership addresses a critical enterprise concern: how to run AI on sensitive data without exposing it to third parties. The SB Energy investment aligns with the growing recognition that power availability is now the binding constraint on AI deployment.
What's next: Whether other component makers join the DSX Ready program and how enterprises adopt confidential computing for AI workloads.
Sources: Yonhap News; ME News; Investing.com.
Nvidia-Backed Armenia Data Center to Reach 300MW with 70,000+ Chips
Firebird's Hrazdan facility is expected to be the largest computing cluster in the region — with export licenses reportedly tied to the Armenia-Azerbaijan peace agreement.
What happened: Firebird, a San Francisco-based startup with Armenian founders, has begun operating a data center in Armenia that is expected to reach 300 megawatts of capacity and more than 70,000 Nvidia chips by the end of next year, according to Bloomberg. One-fifth of the computing power is reserved for domestic needs — local research, government, and the startup ecosystem — while the remainder is designated for foreign companies, including U.S.-based Perplexity AI. The complex in Hrazdan, about 45 kilometers from Yerevan, is expected to be the largest computing cluster in the region and among the largest in Europe. Nvidia chip export licenses were reportedly part of negotiations for the Armenia-Azerbaijan peace agreement.
Why it matters: The Armenia project represents a significant geographic diversification of AI infrastructure beyond the U.S., Europe, and East Asia. It also demonstrates how AI compute is becoming a tool of geopolitical alignment, with export licenses tied to peace negotiations. Armenia, historically dependent on Russia, is using technology investment to strengthen ties with the West.
What's next: Whether the data center reaches its capacity targets and whether other small nations pursue similar AI infrastructure strategies.
Source: Bloomberg.
Palantir Names Nebius Preferred Sovereign AI Infrastructure Partner
The partnership lets commercial customers deploy open models on Nebius infrastructure while retaining control over their data through Palantir's Sovereign AI Operating System.
What happened: Palantir Technologies and Nebius Group announced a partnership to help commercial customers run AI models while retaining control over their computing resources, data, and models. Under the agreement, eligible customers will be able to access Nebius's cloud computing and AI inference services within Palantir's enterprise environment. Customers can deploy open models on Nebius infrastructure and continually adapt them using their own business data. Palantir's Sovereign AI Operating System provides the access controls and isolation needed to work with proprietary data. The partners plan to use modular data centers at locations where power is already available to bring additional computing capacity online faster.
Why it matters: The partnership addresses growing enterprise demand for sovereign AI — the ability to run AI models within an organization's own infrastructure, under its own jurisdiction, with its own data controls. By integrating Nebius's AI-optimized infrastructure with Palantir's access controls, the partnership enables commercial customers who previously lacked a sovereign environment to run their own models.
What's next: Watch for customer adoption metrics and whether other enterprise software vendors announce similar sovereign AI partnerships.
Source: Qazinform.
Cognex to Acquire RealSense for $500 Million.n
The machine vision leader funds the all-cash deal from its balance sheet, strengthening its position in industrial automation and physical AI applications.
What happened: Cognex Corporation, a global leader in industrial machine vision, entered a definitive agreement to acquire RealSense, Inc. for $500 million in cash, funded entirely from existing cash and investments on Cognex's balance sheet. The transaction is expected to close in the fourth quarter of 2026. Before closing, RealSense will spin out its Facial Authentication product line into an independent company. Cognex also expects to provide a three-year $56.5 million cash retention program and to grant restricted stock units valued at approximately $50 million.
Why it matters: The acquisition strengthens Cognex's position in machine vision, a critical technology for industrial automation and robotics. RealSense, formerly Intel's depth-sensing camera division, brings technology used in robotics, drones, and augmented reality — applications that are increasingly relevant as physical AI and autonomous systems expand.
What's next: Whether the deal closes on schedule and how Cognex integrates RealSense's technology into its product portfolio.
Sources: MarketScreener; S&P Capital IQ.
Persistent Systems Acquires Additional 61% Stake in Germany's Nagarro
The deal creates an AI-led digital engineering powerhouse with a combined revenue run rate of approximately $2.9 billion.
What happened: IT services provider Persistent Systems acquired an additional 61.15% stake in German digital engineering firm Nagarro SE, bringing its aggregate shareholding to 83.25% and surpassing the minimum threshold for the transaction to proceed. The takeover offer was executed through Galaxy Germany Holding SE, a wholly-owned subsidiary of Persistent Systems. During the acceptance period ending September 17, 7,568,145 shares representing approximately 61.15% of Nagarro's total share capital were tendered. Under German securities law, an additional two-week acceptance period will run from September 23 to October 6, with the same cash consideration of EUR 81 per share. Persistent plans to pursue a delisting of Nagarro's shares from the Frankfurt Stock Exchange.
Why it matters: The acquisition creates an AI-led digital engineering powerhouse with a combined revenue run rate of approximately $2.9 billion. It reflects the consolidation trend in IT services as companies seek scale and AI capabilities to compete in a rapidly evolving market.
What's next: Whether Persistent completes the remaining steps to close and execute the delisting.
Sources: Moneycontrol; PTI.
Nutanix Acquires Ryax Technologies to Accelerate Agentic AI
The France-based AI platform company expands Nutanix's capabilities in agentic AI and infrastructure automation.
What happened: Nutanix acquired France-based AI platform company Ryax Technologies, expanding its capabilities in agentic AI and infrastructure automation. The acquisition was announced on September 22. Ryax Technologies develops AI platform solutions, and the deal is expected to accelerate Nutanix's agentic AI offerings.
Why it matters: The acquisition reflects growing enterprise demand for AI-powered infrastructure automation, where AI agents manage and optimize hybrid cloud environments. As enterprises deploy AI agents across their infrastructure, the ability to orchestrate and govern those agents becomes a competitive differentiator.
What's next: Watch for integration details and whether Nutanix expands Ryax's capabilities across its hybrid cloud portfolio.
Source: MarketScreener.
Global Power Semiconductor Makers Implement Multiple Price Increases
Infineon, Texas Instruments, STMicroelectronics, and ON Semiconductor raise prices as Taiwan makers plan another 10–15% spot increase in October.
What happened: Global power semiconductor manufacturers, including Infineon, Texas Instruments, STMicroelectronics, and ON Semiconductor, have implemented multiple rounds of price increases in 2026, driven by AI-related demand, according to Securities Times. Taiwan-based power semiconductor makers, including Panjit and Taiwan Semiconductor,r are planning another 10%–15% spot price increase in October. The price increases reflect a structural shift in inventory strategies among overseas IDMs, which are actively holding redundant inventory to hedge against supply gaps. Benefiting from the high-demand cycle, companies including Hua Hong Semiconductor, Dongwei Semiconductor, and Xinlian Integration reported first-half net profit growth exceeding 100%.
Why it matters: Power semiconductors are essential for AI data centers, electric vehicles, and industrial automation, and the sustained price increases signal that demand is outpacing supply across the power electronics supply chain. The multiple rounds of price hikes — rather than a single adjustment — suggest that manufacturers expect tight conditions to persist.
What's next: Whether the price increases continue into 2027 and how they affect the cost structure of AI infrastructure and automotive manufacturers.
Sources: Securities Times; 163.com.
AT&S and Marvell Expand Collaboration for Advanced IC Substrate Capacity
Marvell named as the additional customer at AT&S's Kulim site in Malaysia — addressing a critical but overlooked AI supply chain bottleneck.
What happened: AT&S announced an expanded collaboration with Marvell Technology to increase advanced IC substrate capacity for next-generation AI and cloud infrastructure. Marvell is the additional customer identified in AT&S's previously announced expansion of its Kulim manufacturing site in Malaysia. The long-term agreement builds on the companies' existing relationship and supports Marvell's growing demand for advanced substrates as AI deployments scale. The expansion includes the fit-out of Plant 2 and construction of a new manufacturing facility for IC substrate cores and advanced packaging. The partnership reflects the transition from monolithic chips to chiplet-based designs, which is increasing demand for larger substrate formats, higher integration density, and increased layer counts.
Why it matters: Advanced IC substrates are a critical but often overlooked component of AI infrastructure. As AI chips become larger and more complex, the substrate — the foundation on which chips are mounted — becomes a bottleneck. AT&S's expansion, supported by Marvell's long-term commitment, addresses this constraint and illustrates how the AI supply chain extends beyond wafer fabrication and packaging.
What's next: Whether AT&S's Kulim expansion meets its production timelines and whether other chipmakers sign similar long-term substrate agreements.
Sources: AT&S; Deutsche Börse.
Nasdaq Hits Record High on AI Rally; Micron and SanDisk Surge
Meta's Muse AI assistant drives chip stocks higher while weighing on software and consumer names — Amazon blocked the agent from shopping on its platform.
What happened: The Nasdaq notched a record high close on Tuesday, lifted by Micron Technology and other AI-related stocks, while the S&P 500 hovered just below a record high as oil prices traded around $100 per barrel. Memory chipmakers Micron and SanDisk both rallied. The Philadelphia Semiconductor Index rose more than 2%, with ARM surging 17%, Intel gaining 12%, and AMD rising nearly 10%. The rally was fueled by the strong reception of Meta Platforms' Muse AI assistant, which launched this month and can send emails and complete transactions on behalf of users. However, the success of Muse also weighed on software and consumer stocks, as investors worried it could compete with retailers, online brokerages, and travel platforms. Amazon blocked the agent from shopping on its platform. Uber, Lyft, Expedia, and Charles Schwab all lost ground.
Why it matters: The market's bifurcated reaction — AI chip stocks rallying while software and consumer stocks decline — reflects growing investor conviction that AI agents will disrupt traditional business models even as they drive demand for compute infrastructure. The Muse effect is the clearest example yet of how consumer AI agents could reshape competitive dynamics across multiple sectors.
What's next: Whether the Trump-Xi summit on September 24 produces any breakthroughs on AI regulation or chip export controls, and how software stocks respond to the Muse threat.
Sources: Reuters; Livemint.
ShinyHunters Escalates Campaign with FBI Breach Claim, Clop Infrastructure Hijack
The group demands an eight-figure ransom from the ransomware gang itself — a dramatic reversal in the cybercriminal ecosystem.
What happened: ShinyHunters has claimed a breach of the FBI while simultaneously hijacking the dark web leak site of rival ransomware group Clop, demanding an eight-figure ransom from the ransomware gang itself. The group says it has stolen data on thousands of FBI employees, though the bureau has not commented. Separately, ShinyHunters hijacked Clop's dark web leak site and is demanding a ransom from the ransomware group, a dramatic reversal in the cybercriminal ecosystem.
Why it matters: ShinyHunters' dual targeting of a federal law enforcement agency and a rival ransomware group demonstrates the group's aggressive posture and willingness to operate outside conventional cybercriminal norms. The Clop hijack is particularly notable — it suggests ShinyHunters has compromised Clop's infrastructure, potentially giving it access to Clop's victim data and operations.
What's next: Whether the FBI confirms the breach and how Clop responds to the extortion attempt.
Sources: Reuters.
Ransomware Attacks Target Education and Healthcare Sectors
Booba Project and Qilin claim victims including a high school and a healthcare organization — sectors with sensitive data and limited defenses.
What happened: Ransomware groups including Booba Project and Qilin have claimed attacks on educational and healthcare organizations, continuing a pattern of targeting sectors with sensitive data and limited cybersecurity resources. Tulare Western High School was listed as a victim by the Booba Project group with 35 GB of stolen data. GOTTHELF, a healthcare organization, was also listed by Booba Project with 2 GB of stolen data. Textile City was listed by the Qilin ransomware group.
Why it matters: The attacks illustrate the persistent threat facing educational institutions and healthcare providers, which hold sensitive personal and medical data but often lack the resources for robust cybersecurity defenses. The targeting of a high school is particularly concerning given the vulnerability of student data.
What's next: Whether additional victims are identified and how affected organizations respond.
Source: ransomware. live.
Snorkel AI Valued at $3.5 Billion After $350 Million Raise
The data startup's valuation nearly triples in four months as annualized revenue run-rate crosses $350 million, up from ~$20 million a year earlier.
What happened: Data startup Snorkel AI raised $350 million in fresh funding at a $3.5 billion valuation, nearly tripling its previous valuation of $1.3 billion from May 2025. The round was led by Insight Partners and S32, with participation from existing investors Addition, Greylock, and Wells Fargo. The company's annualized revenue run-rate has crossed $350 million, up from roughly $20 million a year earlier, driven by its data-as-a-service business launched in September 2025. Founded in 2019 by researchers from Stanford's AI lab, Snorkel now offers an "agentic data development platform" that pairs human experts with thousands of specialized AI models to create and vet training data. The company works with frontier AI labs, hyperscalers, enterprises, and the U.S. federal government.
Why it matters: The funding reflects surging demand for high-quality training data as AI labs move beyond simpler labeling work and seek harder, higher-stakes data to train increasingly capable systems. Snorkel's model — selling finished datasets and reinforcement-learning environments rather than charging for human labor — allows it to pay experts more generously while preserving margins.
What's next: Whether Snorkel expands its data-as-a-service offerings and how it competes with rivals including Mercor and Surge AI.
Sources: Reuters; MarketScreener.
AI4S Startup Longxun Kuangteng Raises 100M Yuan Series B1
The capital will fund R&D expansion and a "dry-wet closed-loop" system integrating computational and experimental workflows.
What happened: AI4S (AI for Science) enterprise Longxun Kuangteng completed a 100-million-yuan Series B1 financing round. Long-term shareholder Tongchuang Weiye continued to increase its stake, with Cathay Haitong Kaiyuan and Zhongxin Capital making co-investments. The capital will be used for expanding the R&D team, constructing a dry-wet closed-loop system, and building the commercial operation system.
Why it matters: The funding reflects continued investor interest in AI applications for scientific research, where AI is increasingly being used to accelerate discovery in materials science, drug development, and other fields. The "dry-wet closed-loop" refers to integrating computational (dry) and experimental (wet lab) workflows — a model that is gaining traction in AI-driven scientific discovery.
What's next: Watch for Longxun Kuangteng's research milestones and commercial partnerships.
Source: 36Kr.
VERDA Raises $189 Million for AI Cloud
The funding reflects continued investor appetite for neocloud providers as enterprises and AI labs seek alternatives to the largest hyperscalers.
What happened: AI cloud startup VERDA raised $189 million in a new funding round. The funding reflects continued investor appetite for AI cloud infrastructure, as enterprises and AI labs seek alternatives to the largest hyperscalers.
Why it matters: VERDA joins a growing list of neocloud providers attracting capital to meet surging demand for AI compute.
What's next: Watch for VERDA's customer wins and infrastructure expansion plans.
Sources: Techmeme; The Funding Pulse.
Why It Matters
The AI industry's central contradiction is intensifying. Anthropic and OpenAI released cheaper models just days after their CEOs called for slowing development. OpenAI simultaneously opened its safety evaluation process to third parties while releasing more price-competitive models. The tension between safety rhetoric and commercial competition — driven by pressure from cheaper open-weight models — is becoming the defining dynamic of the frontier AI market. For enterprise buyers, this means the cost of capabilities continues to fall even as governance concerns mount.
AI infrastructure is becoming a geopolitical tool. Nvidia chip export licenses were reportedly part of the Armenia-Azerbaijan peace agreement, and the resulting data center will be the largest computing cluster in the region. This is AI infrastructure as statecraft — a pattern that will likely repeat as smaller nations seek to attract AI investment and align with technology powers.
The cybersecurity threat landscape is escalating in complexity. ShinyHunters claiming a breach of the FBI while simultaneously hijacking Clop's infrastructure represents a new level of audacity in the cybercriminal ecosystem. The group is now targeting both federal law enforcement and rival ransomware operations, demonstrating that the distinction between different types of cybercriminals is blurring.
Memory and power are the AI supply chain's binding constraints. Micron and SanDisk led the semiconductor rally, while Nvidia expanded its ecosystem with LG's cooling and energy storage solutions and increased its SB Energy investment. The market is signaling that memory bandwidth and power availability — not logic chip fabrication — will determine the pace of AI deployment. Nvidia's VAST Data partnership on confidential computing addresses a separate but equally critical constraint: enterprise trust.
The AI trade is decoupling from regulatory uncertainty. Despite the ongoing debate about AI slowdowns and safety, semiconductor stocks rallied to record highs. Investors appear to be differentiating between AI safety rhetoric and actual infrastructure demand. The Muse effect — AI chip stocks up, consumer software stocks down — suggests that markets are pricing in AI's disruptive potential across sectors, not just its infrastructure requirements.
The CODEW Take
Anthropic and OpenAI's cheaper models reveal the real driver of the AI slowdown debate: price competition. The timing is not coincidental. Both labs face intense pressure from cheaper open-weight competitors — including DeepSeek, Zhipu, and others — that are offering comparable capabilities at a fraction of the cost. Amodei's slowdown essay and OpenAI's third-party safety evaluation framework serve multiple purposes: they position the labs as responsible actors ahead of regulation, create cover for delayed IPOs, and shift the competitive narrative from raw capability to governance and trust. But the simultaneous release of cheaper models reveals the commercial reality: frontier labs must compete on price and efficiency, not just capability. For enterprise buyers, this is good news — the cost of AI is falling. For the labs, it means the slowdown debate is as much about managing investor expectations as it is about safety.
The FBI breach claim, even if unverified, is a significant escalation. ShinyHunters' claim that it breached the FBI — and its simultaneous hijacking of Clop's dark web infrastructure — demonstrates a group operating with unprecedented audacity. Whether or not the FBI breach is confirmed, the pattern is clear: ShinyHunters is targeting the highest-profile organizations and rival cybercriminal groups alike. The hijacking of Clop's leak site is particularly notable because it suggests ShinyHunters has compromised Clop's operational infrastructure, potentially giving it access to victim data and intelligence. This is not just a breach — it's a takeover.
Nvidia's ecosystem strategy is becoming its most important competitive moat. The DSX Ready certifications for LG's cooling and energy storage solutions, the VAST Data confidential computing partnership, and the SB Energy investment collectively demonstrate that Nvidia is building an ecosystem that extends far beyond chips. Each partnership addresses a specific bottleneck: cooling, power, data management, enterprise trust. By orchestrating these components, Nvidia makes itself indispensable not just as a chip supplier but as the platform for AI infrastructure. Competitors can build faster chips, but replicating Nvidia's ecosystem is a much harder problem.
Armenia's data center is a preview of AI infrastructure's next phase. The 300MW, 70,000-chip facility in Hrazdan is significant not just for its scale but for its location. Armenia — historically dependent on Russia — is using AI infrastructure investment to strengthen ties with the West. The fact that Nvidia export licenses were part of the peace agreement negotiations shows that AI compute has become a diplomatic tool. Expect more small nations to pursue similar strategies, trading geopolitical alignment for AI infrastructure investment.
The Muse effect is the clearest signal yet that AI agents will reshape multiple industries. Meta's Muse AI assistant — which can send emails, book travel, and complete purchases — has driven AI chip stocks higher while simultaneously weighing on consumer software stocks including Uber, Lyft, Expedia, and Charles Schwab. Investors are pricing in a world where AI agents disintermediate traditional platforms. Amazon's decision to block Muse from shopping on its platform shows that incumbents are already defending their turf. The AI agent economy is arriving, and it will create both winners and losers across the consumer technology landscape.
What to Watch
- Trump-Xi summit on September 24: AI regulation, chip export controls, and critical minerals are on the agenda — could move semiconductor stocks.
- OpenAI's third-party evaluation framework: How regulators and independent safety organizations respond to the company's plan to involve external evaluators earlier in the development process.
- FBI breach claim: Whether the FBI confirms the ShinyHunters breach and what data was actually compromised.
- Nvidia's DSX Ready ecosystem expansion: Whether other component makers join the program and how enterprises adopt confidential computing.
- Armenia data center progress: Whether Firebird reaches its 300MW, 70,000-chip target by the end of 2027.
- Persistent Systems-Nagarro delisting: Whether the delisting from the Frankfurt Stock Exchange proceeds as planned.
- Snorkel AI's growth trajectory: Whether the company sustains its $350 million revenue run-rate and expands its customer base.
- Power semiconductor price increases: Whether the multiple rounds of price hikes continue into 2027.
Editorial Note
The CODEW Daily News Coverage tracks the most important technology developments of the day, with a focus on AI, enterprise software, cloud computing, semiconductors, cybersecurity, startups, and digital infrastructure. This edition is a pure news aggregation and coverage page — what happened — while the Daily Tech Briefing provides the deeper strategic interpretation.
Coverage is based on company announcements, public disclosures, industry reporting, and other publicly available information. Reported figures and sourced-but-unconfirmed details are noted as such.
Reviewed by Erwin Castro
on
Wednesday, September 23, 2026
Rating:
