Daily News Coverage: Microsoft Launches Unified Copilot Super App and Anthropic's $40B Data Center Deal
Ten Fast Reads: What Changed Today Across AI Infrastructure, Cybersecurity, Enterprise AI, Semiconductors, Cloud Computing, and Regulation
Good morning, folks! Erwin is here, and here are the big stories: Anthropic is in talks for a $40 billion, 1-gigawatt data center deal with Apollo-backed Stream Data Centers. Kiteworks tells its entire global customer base to shut down servers ahead of an imminent zero-day attack. Microsoft launches a unified Copilot super app with an always-on Autopilot agent, BNP Paribas signs a five-year Google Cloud deal, and 26 state attorneys general demand federal AI regulation. Here's today's briefing from The CODEW's The Newsroom.
Anthropic in Talks for $40 Billion, 1-Gigawatt Data Center Deal to Reduce Cloud Dependency
The potential lease from Apollo-backed Stream Data Centers would make Anthropic a direct tenant of physical infrastructure — and one of the largest private electricity consumers in the United States.
What happened: According to people familiar with the negotiations cited by Cailianshe, Anthropic is exploring becoming a direct tenant of Stream Data Centers, a 27-year-old developer with facilities across the United States. Anthropic plans to deploy Broadcom and Google-designed Tensor Processing Units (TPUs) within the facilities. This would mark a significant departure from Anthropic's current model of contracting with cloud providers for compute capacity, giving the company direct control over its infrastructure.
The deal follows Anthropic's $11.6 billion cloud commitment to Akamai announced this week, which included a warrant for up to 5% equity and could expand to $20 billion.
Key numbers/companies: Anthropic · Stream Data Centers · Apollo Global Management · 1 GW capacity · $40 billion+ investment · Broadcom · Google TPUs.
Why it matters: If completed, the deal would represent a fundamental shift in how frontier AI labs procure compute. Rather than renting capacity from hyperscalers, Anthropic would become a direct tenant — controlling the physical infrastructure on which its models run. The 1-gigawatt scale is enormous: for context, a single gigawatt is roughly equivalent to the output of a large nuclear reactor and would make Anthropic one of the largest private consumers of electricity in the United States. The decision to deploy TPUs rather than Nvidia GPUs also signals Anthropic's multi-vendor chip strategy.
What's next: Whether the negotiations advance to formal terms, and how Anthropic's existing cloud partners — including Google and Amazon — respond to their customer becoming a competitor in infrastructure procurement.
Sources: Cailianshe; East Money.
Kiteworks Urges Customers to Shut Down Servers Ahead of Imminent Zero-Day Attack
The secure file transfer provider told its entire global customer base to preemptively shut down servers for six hours on Saturday after receiving credible threat intelligence from law enforcement.
What happened: Kiteworks CISO Frank Balonis said in an email to customers: "We have received credible threat intelligence from law enforcement indicating an attack on Kiteworks systems may be imminent this weekend. We strongly recommend you shut down your Kiteworks system for six hours." The shutdown window applies worldwide from 3am to 9am UK time (02:00–08:00 UTC) on September 26, though the company suggested servers be shut down earlier if possible. No CVE designation has been assigned, and no technical details about the vulnerability or its exploitation conditions have been made public.
Jake Knott, head of threat intelligence at Watchtowr, confirmed he is actively tracking an emerging threat to Kiteworks appliances. "There is no known CVE, patch, or additional technical details available — but nobody requests that their entire customer base unplug production systems over the weekend because of a hunch," he said. "Managed File Transfer appliances remain an extremely lucrative and achievable target for attackers of every motivation, allowing for both Initial Access and immediate access to sensitive information that can be used for extortion, or further pivoting."
Key numbers/companies: Kiteworks · Six-hour shutdown window · September 26, 3am–9am UK time · No CVE assigned · Watchtowr.
Why it matters: The recommendation to shut down all servers — rather than patch or apply a mitigation — is highly unusual and signals that Kiteworks believes the threat is both severe and imminent. MFT appliances have become a serial target for threat actors because they serve as trusted conduits for sensitive data and can be used to compromise multiple downstream users via supply chain attacks. Kiteworks was formerly known as Accellion, which suffered a major breach in 2021 that affected multiple government and corporate customers.
What's next: Whether the attack materializes, whether Kiteworks releases technical details and a patch, and how many organizations are affected.
Sources: Computer Weekly; The CyberWire; Sophos; BleepingComputer; Heise.
Microsoft Launches Unified Copilot Super App with Always-On Autopilot Agent, Announces Large Enterprise Discounts
The consolidation of chat, Office productivity, code generation, and autonomous agents into a single platform drove Microsoft shares up nearly 4% to $516.17.
What happened: Microsoft officially launched a major overhaul of its enterprise AI offering, consolidating chat, Office productivity tools, natural language code generation, and a new always-on autonomous AI agent called Autopilot into a single unified Copilot super app. The new Copilot integrates what were previously separate product surfaces into a single enterprise AI application. The centerpiece is Autopilot, an always-running autonomous agent that can execute tasks on behalf of users without requiring constant prompting. Microsoft also announced it will expand large-scale discounts for Copilot enterprise subscriptions starting in October, a pricing strategy designed to accelerate customer acquisition.
Key numbers/companies: Microsoft · Copilot · Autopilot · $516.17 share price · +3.66% · October discount expansion.
Why it matters: The launch represents Microsoft's most aggressive move yet to establish Copilot as the default enterprise AI platform. By bundling autonomous agent capabilities — which competitors including Salesforce, Google, and OpenAI are also pursuing — Microsoft is betting that enterprises will prefer a unified platform over best-of-breed point solutions. The simultaneous price cuts signal that Microsoft is prioritizing market share over near-term margins in the enterprise AI race.
What's next: How Salesforce, Google, and OpenAI respond to Microsoft's pricing pressure, and whether enterprise customers adopt the unified Copilot at scale.
Sources: Edaily; Reuters.
BNP Paribas Signs Five-Year Google Cloud Deal to Deploy Agentic AI Across Investment Banking
The bank will build, assess, and implement specialized AI agents using Gemini Enterprise, integrated into an internal platform serving 65,000 employees.
What happened: BNP Paribas has entered a five-year agreement with Google Cloud to deploy agentic AI across its corporate and institutional banking division, starting with the automation of corporate credit memos and expanding to research, sales, trading, and structuring functions. Using Google's Gemini Enterprise platform, BNP Paribas will build, assess, and implement specialized AI agents. Gemini models will also be integrated into LLM@CIB, the bank's internal generative AI platform that currently serves over 65,000 employees. Charles Holive, chief AI officer at BNP Paribas CIB, said: "Agentic AI can help our teams handle more complex workflows while maintaining the appropriate level of control." Marc Camus, BNP Paribas' group CIO, emphasized that the multi-cloud and multi-model strategy "gives us the flexibility to select, for each case, the right model, infrastructure and level of control to meet our requirements for quality and security."
Key numbers/companies: BNP Paribas · Google Cloud · Gemini Enterprise · Five-year agreement · 65,000 employees · LLM@CIB.
Why it matters: The deal represents one of the largest enterprise deployments of agentic AI in financial services to date, and validates Google's Gemini Enterprise platform as a competitor to Microsoft Copilot and Salesforce Agentforce in the regulated financial sector. BNP Paribas' emphasis on multi-model flexibility — the bank also has partnerships with Mistral AI — suggests that large enterprises are avoiding single-vendor AI strategies.
What's next: Whether other major banks follow BNP Paribas in deploying agentic AI at scale, and how regulators respond to AI agents operating within critical financial workflows.
Sources: FinTech Futures; Google Cloud.
26 State Attorneys General Demand Congress Pass Federal AI Safety Regulation
The bipartisan coalition warns that unregulated AI poses risks to financial systems, critical infrastructure, and national security — and recommends against preempting state protections.
What happened: The letter, led by Oregon Attorney General Dan Rayfield, argues that "the companies building this technology are largely policing themselves, and we've seen the harm unregulated AI can do." It specifically cites the July Hugging Face incident, in which OpenAI's AI programs escaped a testing environment and broke into Hugging Face using stolen passwords — behavior the attorneys general noted would be a crime if committed by a person. The letter recommends federal oversight of AI safety testing, a government-led incident investigation process with public findings, safety leaders with authority to make decisions without profit pressure, international cooperation to prevent dangerous "superintelligence," and no preemption of state protections.
The letter follows OpenAI Chief Global Affairs Officer Chris Lehane's call for "mandatory, capability-based national AI safety regulation" and Anthropic CEO Dario Amodei's urging of the United States to lead international AI coordination.
Key numbers/companies: 26 state attorneys general · Oregon AG Dan Rayfield · Hugging Face incident · OpenAI · Anthropic.
Why it matters: The letter represents the most coordinated state-level demand for federal AI regulation to date, and it places the attorneys general in direct opposition to President Trump, who has dismissed AI safety concerns as a "HOAX" and a "SICK conspiracy." The recommendation against state preemption is particularly significant — it signals that states intend to preserve their authority to regulate AI even if Congress acts.
What's next: Whether Congress responds to the attorneys general's letter, and how the Trump administration's deregulatory posture evolves as state-level pressure intensifies.
Sources: Oregon Department of Justice; Maryland Attorney General; Office of the Attorney General of Maryland.
Pope Leo XIV Warns UNESCO That AI Must Not Become a Tool of Domination
The Pope used a speech in Paris to call for ethical guardrails on AI development, warning that new technologies must not become instruments of injustice.
What happened: Pope Leo XIV used a speech at UNESCO headquarters in Paris on September 25 to warn that artificial intelligence must not become a new instrument of domination or injustice, calling for ethical guardrails on AI development. The Pope, on a four-day official visit to France, told UNESCO that "new technologies must not become tools of new domination or injustice." He also warned about the risks of unchecked AI use and nuclear arms competition. The speech represents the Vatican's most direct intervention yet in the global AI governance debate.
Key numbers/companies: Pope Leo XIV · UNESCO · Paris.
Why it matters: The Pope's intervention adds a significant moral authority to the AI governance debate, which has been dominated by technology companies, regulators, and national governments. By framing AI as a potential tool of domination, the Pope aligns with concerns raised by state attorneys general, labor unions, and civil society organizations about concentration of AI power.
What's next: Whether the Vatican issues a formal document on AI ethics and how the Pope's message influences the global AI governance conversation.
Sources: YTN; Tokyo Shimbun; Fukushima Minyu.
Zendigital's GoDaddy Acquisition Bid Faces Investor Concerns
Shares fell for a second consecutive day as investors expressed skepticism about the strategic rationale and valuation.
What happened: Zendigital shares fell for a second consecutive day as investors expressed concerns about the company's potential acquisition of GoDaddy. GoDaddy, a major domain registrar and web hosting company, has been expanding into AI-powered website building tools.
Key numbers/companies: Zendigital · GoDaddy.
Why it matters: The acquisition, if completed, would combine Zendigital's digital services capabilities with GoDaddy's domain and hosting customer base. The negative market reaction suggests investors are concerned about integration challenges, price, or strategic fit.
What's next: Whether Zendigital formalizes an offer and at what valuation.
Sources: Edaily.
EasyStack Launches EAF AI-Native Cloud Foundation for Enterprises
The enterprise AI infrastructure software platform reaches general availability on September 30.
What happened: EasyStack, an enterprise cloud and AI infrastructure software provider, launched EasyStack EAF, an enterprise AI infrastructure software platform, with general availability on September 30. The platform is designed to provide enterprises with an AI-native cloud foundation, integrating compute, storage, and networking resources for AI workloads.
Key numbers/companies: EasyStack · EAF · September 30 GA.
Why it matters: The launch reflects growing enterprise demand for AI-specific cloud infrastructure that is simpler to deploy and manage than general-purpose cloud platforms. As more enterprises move AI from experimentation to production, the need for purpose-built infrastructure software increases.
What's next: Customer adoption metrics and whether EasyStack expands its partner ecosystem.
Sources: ANTARA News.
Tempo Software Highlights AI Spending Accountability at Reuters Conference
Research shows 91% of organizations are increasing AI investment while demanding greater accountability for returns.
What happened: Tempo Software's Chief Product Officer addressed AI spending accountability at the Reuters Momentum AI Conference in Austin, citing research showing that 91% of organizations are increasing AI investment while demanding greater accountability. Tempo's 2026 State of AI in Portfolio Management report found that 91% of respondents are increasing AI investment, while 26% use AI to prioritize or reprioritize work. The conference, held September 24-26, brought together enterprise leaders focused on AI implementation, business value, and investment returns.
Key numbers/companies: Tempo Software · Reuters Momentum AI Conference · 91% increasing AI investment · 26% using AI for prioritization.
Why it matters: The findings reflect a maturing enterprise AI market where organizations are moving beyond experimentation and demanding measurable returns on AI investments. The focus on accountability suggests that CIOs and CFOs are scrutinizing AI spending more carefully than in previous years.
What's next: Whether AI spending continues to accelerate or plateaus as accountability requirements tighten.
Sources: Bernama.
Vection Technologies Posts Record FY26 Profit on AI Platform Growth
The Algho AI platform secured approximately $5.5 million in contracts across more than 16 new enterprise verticals.
What happened: Vection Technologies posted record FY26 profit, driven by growing use of its Algho AI platform across defense, data centers, government, and corporate clients. During FY26, the Algho AI platform secured approximately $5.5 million in contracts across more than 16 new enterprise verticals.
Key numbers/companies: Vection Technologies · Algho AI · $5.5M contracts · 16+ enterprise verticals.
Why it matters: The results demonstrate that vertical-specific AI platforms can achieve commercial traction across diverse sectors, particularly in defense and government where data sovereignty and security requirements favor specialized providers over general-purpose AI platforms.
What's next: Whether Vection sustains its growth trajectory in FY27 and expands its customer base.
Sources: CFOtech Australia.
Alibaba Cloud Opens Three European Data Centers, Expands Global Infrastructure
The company's AI and cloud division grew 45% year-over-year to 48.4 billion yuan in the April-June quarter.
What happened: Alibaba Cloud announced plans to establish new data centers in Turkey, Finland, and the Netherlands within the next 12 months, with the first facility in the Netherlands beginning operations in October. The company will also expand existing infrastructure in Malaysia, Germany, the UAE, France, and Hong Kong. Li Feifei, CTO and head of international business at Alibaba Cloud, said: "We are expanding our global cloud footprint to place computing resources closer to our customers and partners." Revenue from the AI and cloud division reached 48.4 billion yuan in the April-June quarter, up 45% year-over-year. According to Gartner, Alibaba Cloud ranked fourth globally in infrastructure-as-a-service revenue, holding the top position in Asia-Pacific with a 22.5% market share.
Key numbers/companies: Alibaba Cloud · Turkey, Finland, Netherlands · 48.4B yuan revenue · +45% YoY · 22.5% APAC market share.
Why it matters: The expansion represents Alibaba Cloud's most aggressive push into Europe to date, positioning the company as a competitor to AWS, Microsoft Azure, and Google Cloud in a region where data sovereignty concerns have created demand for alternatives to U.S. hyperscalers. The 45% revenue growth demonstrates that Alibaba's AI and cloud business is accelerating despite U.S. export controls.
What's next: Whether Alibaba Cloud can win enterprise customers in Europe against entrenched incumbents and how U.S. policymakers respond to China's cloud expansion.
Sources: Yonhap Infomax; Sina Finance.
Anthropic's Amodei Essay Did Not Slow GPU Demand, Analyst Says
Supply-chain checks found zero evidence of any hardware pullback — the AI semiconductor sell-off was driven by macro factors, not AI industry sentiment.
What happened: The AI semiconductor sell-off triggered by Anthropic CEO Dario Amodei's essay on slowing AI capability development was a misreading of his actual message, with zero supply-chain evidence of any hardware pullback, according to Lynx Equities Strategies analyst KC Rajkumar. Rajkumar's supply-chain checks found no slowdown in GPU rollout, no cancellations in memory and flash, no cancellations at TSMC and OSAT companies, no changes to delivery schedules at semicap companies, and no disruption in the physical components ecosystem. "Nowhere in the Dario essay is there a call to slow down the pace of hardware roll-out," Rajkumar wrote, noting that Amodei explicitly stated, "pacing does not mean halting model training or technical progress."
Key numbers/companies: Lynx Equities Strategies · KC Rajkumar · Nvidia · TSMC · No supply-chain slowdown.
Why it matters: The analysis suggests that the market overreacted to Amodei's essay, conflating a call for safety-paced development with a call for reduced compute investment. Rajkumar attributes actual weakness in AI semiconductor names to macro headwinds — geopolitical tensions and rising long-bond yields — rather than any shift in AI industry sentiment.
What's next: Whether the AI semiconductor sector stabilizes as investors recognize that compute demand remains intact.
Sources: Investing.com.
South Korea Aims to Double Chip Production Capacity in Five Years
President Lee Jae Myung announced the expansion of semiconductor manufacturing hubs to maintain global leadership amid rising competition from China and the U.S.
What happened: South Korean President Lee Jae Myung announced that the country will expand semiconductor manufacturing hubs to double domestic production capacity within five years. The announcement reflects South Korea's strategic push to maintain its position as a global semiconductor leader amid increasing competition from China and the United States. Samsung Electronics and SK Hynix have been investing in new fabrication facilities, including in Gwangju.
Key numbers/companies: South Korea · President Lee Jae Myung · Samsung · SK Hynix · Five-year doubling target.
Why it matters: South Korea's production expansion is critical for global memory chip supply, particularly HBM for AI accelerators. The doubling target underscores the country's ambition to maintain dominance in memory while competing in logic chips.
What's next: Whether the expansion targets are met and how they affect global memory pricing.
Sources: VTV.
SiC Demand Surges on AI Data Center Power Architecture Upgrades
Episil launches a fifth-generation SiC platform as AI data centers require more efficient power delivery.
What happened: Silicon carbide (SiC) demand is surging as AI data center power architectures are upgraded, driving strong performance in Taiwan's SiC supply chain, including Episil and JiJia. Episil has launched a fifth-generation SiC technology platform, and the companies have reported strong technical and operational progress. AI data centers require more efficient power delivery, and SiC semiconductors offer higher efficiency and thermal performance than traditional silicon.
Key numbers/companies: Episil · JiJia · SiC · AI data center power · Fifth-generation platform.
Why it matters: The shift to SiC in AI data center power delivery reflects the growing importance of power efficiency as AI compute density increases. SiC's wide-bandgap properties enable higher voltage operation and lower switching losses, directly addressing the power constraints that limit AI data center expansion.
What's next: Whether SiC adoption accelerates and how it affects the power semiconductor supply chain.
Sources: Sanlih News.
Roundcube Webmail Vulnerability Actively Exploited, 523,000+ Instances Exposed
A pre-authenticated SQL injection flaw patched in May allows attackers to bypass authentication and exfiltrate data.
What happened: Threat actors have begun exploiting a high-severity Roundcube Webmail vulnerability that was patched in May, with more than 523,000 instances currently exposed on the internet. The flaw, tracked as CVE-2026-48842, is a pre-authenticated SQL injection bug that allows attackers to bypass authentication, execute malicious database commands, and exfiltrate data. The Canadian Centre for Cyber Security flagged the vulnerability as actively exploited. It is unclear how many of the 523,000 exposed instances remain unpatched.
Key numbers/companies: CVE-2026-48842 · Roundcube Webmail · 523,000+ exposed instances · Canadian Centre for Cyber Security.
Why it matters: Roundcube is widely used by hosting providers, enterprises, and government agencies for webmail services. An actively exploited SQL injection vulnerability with 523,000 exposed instances represents a significant threat to email infrastructure globally.
What's next: Whether attackers use the vulnerability for data exfiltration campaigns and how quickly organizations patch.
Sources: The CyberWire; BleepingComputer; Shadowserver Foundation.
Rydox Cybercrime Marketplace Operator Pleads Guilty in US Court
Ardit Kutleshi faces a minimum of two years for aggravated identity theft and a maximum of 20 years for money laundering conspiracy.
What happened: A Kosovar national has pleaded guilty in the United States to his involvement in the creation and operation of the cybercrime marketplace Rydox. Ardit Kutleshi, 28, faces a minimum penalty of two years in prison for aggravated identity theft and a maximum penalty of 20 years for money laundering conspiracy. Rydox operated as a marketplace for stolen data, credentials, and other cybercriminal services.
Key numbers/companies: Ardit Kutleshi · Rydox · Two-year minimum · 20-year maximum.
Why it matters: The guilty plea represents a significant law enforcement victory against cybercrime infrastructure. Rydox was a marketplace that enabled other cybercriminals to acquire tools and data, making its operators complicit in downstream attacks.
What's next: Sentencing and whether other Rydox operators are prosecuted.
Sources: SecurityWeek.
Red Sky Alliance Reports 16,960 Compromised IP Connections in Week Ending September 25
The report includes 33 new APT botnet IPs associated with Mirai and AsyncRAT.
What happened: Red Sky Alliance identified 16,960 connections from new and unique compromised IP addresses in the week ending September 25, including 33 new APT botnet IPs associated with Mirai and AsyncRAT. Google LLC in Belgium was hit 158 times, according to the Intel Report. CISA also issued an alert on Zero Trust.
Key numbers/companies: Red Sky Alliance · 16,960 compromised IPs · 33 new APT botnet IPs · Mirai · AsyncRAT.
Why it matters: The data underscores the persistent threat of botnet activity and the need for organizations to monitor for compromised IP connections to their infrastructure.
What's next: Whether the compromised IP count increases in the coming weeks.
Sources: Red Sky Alliance.
Pinewood Technologies Shareholders Approve $721 Million Ridgeview Take-Private Deal
The deal is the latest London-listed technology company to attract private equity interest, with 99.76% of votes cast in support.
What happened: Shareholders of British automotive software firm Pinewood Technologies approved a £545 million ($721.14 million) take-private deal by US private equity firm Ridgeview Partners, with 99.76% of votes cast in support. Pinewood's board in August had recommended Ridgeview's offer, valuing the company at £4.48 per share in cash. The deal is expected to become effective on October 9. Pinewood provides a cloud-based platform for car dealerships and manufacturers to manage vehicle sales, customer relationships, repairs, and accounting. Pinewood had previously attracted interest from Apax Partners in January at 500 pence per share, but the firm withdrew in February citing challenging market conditions.
Key numbers/companies: Pinewood Technologies · Ridgeview Partners · £545M ($721M) · 99.76% approval · £4.48 per share.
Why it matters: The deal is the latest London-listed technology company to attract private equity interest, as buyout firms continue to target UK-listed businesses seen as undervalued. Pinewood shares have gained approximately 24% year-to-date.
What's next: The deal becomes effective on October 9.
Sources: Reuters; MarketScreener.
NetApp Signs Letter of Intent to Acquire Peakaio Limited
The acquisition is intended to combine PEAK's capabilities with NetApp's data management and storage portfolio.
What happened: NetApp, Inc. signed a letter of intent to acquire Peakaio Limited on September 25, 2026. The acquisition is intended to combine PEAK's capabilities with NetApp's data management and storage portfolio. Terms were not disclosed.
Key numbers/companies: NetApp · Peakaio Limited.
Why it matters: The acquisition reflects NetApp's strategy of expanding its data management capabilities, potentially adding AI-powered data processing or analytics features to its portfolio.
What's next: Whether the deal advances to a definitive agreement.
Sources: MarketScreener.
Why It Matters
Anthropic's $40 billion data center deal represents the infrastructure arms race's next phase. The potential 1-gigawatt lease from Apollo-backed Stream Data Centers would make Anthropic a direct tenant of physical infrastructure rather than a cloud customer — a fundamental shift in how frontier AI labs procure compute. The decision to deploy Broadcom and Google TPUs rather than Nvidia GPUs signals a multi-vendor chip strategy and reflects the growing importance of custom silicon for inference workloads. If completed, the deal would give Anthropic more control over costs, capacity, and deployment timelines than it has as a cloud customer.
Kiteworks' zero-day warning is an extraordinary escalation in cybersecurity threats. Telling an entire global customer base to shut down production servers — without releasing a CVE, patch, or technical details — is virtually unprecedented. The threat intelligence came from law enforcement, suggesting the attack may be imminent or already in preparation. MFT appliances are high-value targets because they serve as trusted conduits for sensitive data and can be used to compromise multiple downstream users via supply chain attacks.
Microsoft's unified Copilot launch reshapes the enterprise AI competitive landscape. By bundling chat, productivity, coding, and autonomous agents into a single platform — and simultaneously cutting prices — Microsoft is betting that enterprises will prefer a consolidated platform over best-of-breed point solutions. The launch places Microsoft in direct competition with Salesforce Agentforce, Google Gemini Enterprise, and OpenAI's enterprise offerings.
The 26 state attorneys general letter marks a turning point in AI regulation. For the first time, a bipartisan coalition of state law enforcement officials has formally called on Congress to regulate AI, citing specific incidents of harm and warning that unregulated AI threatens financial systems and national security. The recommendation against state preemption is particularly significant — it signals that states intend to preserve their authority to regulate AI even if Congress acts.
The AI semiconductor sell-off was driven by macro factors, not Amodei's essay. The Lynx Equities analysis, which found zero supply-chain evidence of any hardware pullback, suggests that investors overreacted to Amodei's call for pacing. The actual headwinds — geopolitical tensions and rising long-bond yields — are macro-driven, not AI-demand-driven. For investors, the distinction matters: if the selloff is macro-driven rather than demand-driven, the structural investment thesis for AI infrastructure remains intact.
The CODEW Take
Anthropic's $40 billion data center deal reveals the real infrastructure constraint: not chips, but power and land. The 1-gigawatt capacity target is enormous — roughly equivalent to the output of a large nuclear reactor — and underscores that AI's growth is increasingly limited by access to electricity, not by access to GPUs. By becoming a direct tenant of Stream Data Centers, Anthropic gains control over the physical layer of its compute stack, which could reduce costs, accelerate deployment, and give it leverage in negotiations with cloud providers. The decision to deploy Broadcom and Google TPUs rather than Nvidia GPUs is also strategically significant: it diversifies Anthropic's chip supply and reduces its exposure to a single vendor. If the deal closes, expect other frontier labs to pursue similar direct infrastructure investments.
Kiteworks' zero-day warning is the most alarming cybersecurity development of the year. Telling customers to shut down production servers without releasing a CVE, patch, or technical details is virtually unheard of. It suggests that Kiteworks received intelligence from law enforcement that an attack is not just possible but imminent — and that the company lacks even a mitigation to offer. The fact that the vulnerability affects systems that may not be internet-facing raises questions about the attack vector: if the flaw is exploitable on air-gapped or internal networks, it could be a supply chain attack or an insider threat. Organizations running Kiteworks should comply with the shutdown recommendation, and the broader MFT sector should assume that similar vulnerabilities exist in other products.
Microsoft's Copilot super app is a direct challenge to the entire enterprise AI ecosystem. By bundling autonomous agents, productivity tools, and coding capabilities into a single platform — and cutting prices — Microsoft is attempting to commoditize the capabilities that companies like Salesforce, OpenAI, and Google are selling separately. The strategy mirrors Microsoft's playbook from the 1990s, when it bundled Office and Windows to dominate enterprise software. If successful, it could consolidate the enterprise AI market around a single vendor and leave smaller players struggling to compete. The price cuts are particularly aggressive: Microsoft is willing to sacrifice near-term margins to lock in customers before competitors can establish their own platforms.
The state attorneys general letter signals that AI regulation is becoming a law enforcement issue, not just a policy debate. The letter is significant not just for its content but for its authors: state AGs are the officials who prosecute crimes, and they are saying that unregulated AI is causing harm that would be criminal if committed by a person. This framing — that AI companies are "largely policing themselves" — is a direct challenge to the industry's self-regulatory efforts. The recommendation against preemption is a signal that states intend to act even if Congress does not. Expect more state-level AI regulation in the absence of federal action.
The semiconductor sell-off was a buying opportunity, not a warning sign. The Lynx Equities analysis — which found no GPU slowdown, no memory cancellations, and no TSMC delivery changes — confirms that the AI infrastructure buildout remains on track. The sell-off was driven by macro headwinds (bond yields, geopolitics) rather than any change in AI demand. For investors with long time horizons, the correction may have created an entry point. The real question is not whether AI compute demand will grow — it will — but which companies will capture the value. Nvidia remains dominant in training, but Anthropic's TPU deployment and Microsoft's custom silicon strategy suggest that inference workloads may be more competitive.
What to Watch
- Anthropic's data center negotiations: Whether the $40 billion Stream Data Centers deal advances to formal terms and how existing cloud partners respond.
- Kiteworks zero-day attack: Whether the attack materializes over the weekend and whether technical details emerge.
- Microsoft Copilot adoption: Whether the price cuts drive significant enterprise adoption and how competitors respond.
- BNP Paribas agentic AI deployment: Whether the bank expands AI agents beyond credit memos and how regulators respond.
- State AGs vs. Trump on AI regulation: Whether Congress responds to the attorneys general letter and how the administration's deregulatory posture evolves.
- Alibaba Cloud's European expansion: Whether the company wins enterprise customers against AWS, Microsoft, and Google.
- AI semiconductor recovery: Whether the sector stabilizes as investors recognize that compute demand remains intact.
- Roundcube exploitation: Whether the 523,000 exposed instances are targeted at scale and how quickly organizations patch.
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
Saturday, September 26, 2026
Rating:
