How is Accenture managing employee spend on AI tools like Claude?

Accenture is setting employee spend to approximately $30 a month per person on Claude, with larger projects receiving specific token allocations that translate into dollar spend.

What custom silicon did Microsoft announce for its AI infrastructure?

Microsoft announced the Maia 300, with multi-gigawatt deployments starting next year.

Which company is acquiring OnLogic and Interlogic, and for what purpose?

Omron is acquiring OnLogic and Interlogic for $675 million to grow its physical AI portfolio.

Jay Cuthrell October 18, 2026 WTF

What The Fudge for October 18, 2026: AI in the Workplace, Security Paradoxes, Funding, and M&A

Enterprise AI spend limits, IoT and frontier model security paradoxes, custom silicon and fab timelines, AI startup funding rounds, and strategic M&A.

Subscribe and Listen: https://youtu.be/ruR4YnwCo6c

🎙️ The “Rough Topics” Edition

Welcome to What The Fudge for Sunday, October 18, 2026. This week, we are trying another format. Instead of a specific company breakout, we are organizing by rough topics: AI in the workplace, AI security paradoxes, AI infrastructure, venture capital funding updates, and merger and acquisition updates.

💼 AI and the Workplace

If you are consuming AI in your current workplace, there is a very good chance that you have been put on some kind of spend limit, allowance, or threshold.

Enterprise Spend Thresholds: Whenever Accenture talks about earnings, related operational stories tend to surface. Across AI tools, tracking employee spend can easily reach into the thousands of dollars a month per employee depending on the use case, especially with Anthropic Claude if you are using coding tools or high-end, set-and-forget features. At Accenture, employees will now spend ~$30 a month per person on Claude. Given the size of Accenture’s staff, that aggregate number gets very large, very quickly. Presumably, $30 a month per person is an average, with larger projects receiving specific token allocations that translate into dollar spend. So, I would gladly pay you Tuesday for a GPU today?
Addressing Code Backlogs with Contingent Labor: Looking at financial institutions, Bank of America is hiring an army of contractors to begin patching software based on AI findings from models like Mythos. When looking back at internally developed codebases, whether addressing a bug, a defect, a functional backlog, or security analysis, there is not necessarily a third-party vendor to rely on. Using contingent labor contractors provides a structured way to address and sunset those identified issues.
Internal Productivity and Societal Shifts: Anthropic released an internal hiring and productivity report. Of course, this report is a bit self-serving since Anthropic literally markets and sells the tools enabling what they claim, a classic case of “look how tasty our dog food is, look how tasty our champagne is, and imagine what it might be for you.” Beyond pure productivity metrics, it is also worth examining the broader human dynamic, as covered in a recent interview of MIT sociologist Sherry Turkle.
Machine Learning in Live Audio: AI tools are proliferating beyond knowledge work. In the world of machine learning and live audio, L-Acoustics provides voice isolation for live performances. Similar to noise cancellation in headsets or a radio squelch knob, these algorithms cut through background noise or chatter. Whether it is voice isolation or guitar isolation, a new generation of performers looking to improve their stagecraft will increasingly use neural DSP in their custom in-ear monitor blends to get the best possible feel and flow on stage.
Platform and Data Enrichment: On the platform front, Figma has defined how AI agents work on its platform as a domain specific language. Meanwhile, Bloomberg is adding conversational AI to all Terminal users to enable muxing together internal content with financial data.

🔐 The AI Security Paradox

The nature of AI generates clear security paradoxes: AI can make defensive security better, including claims of solving long-standing cryptographic questions, such as the so-called NSA backdoor while simultaneously introducing new vectors for cyberattacks and operational vulnerabilities.

Self-Replicating Models and the Internet of Things: We have long asked why a refrigerator or toaster oven needs to be connected to the internet. Anthropic now belives self-replicating IoT devices are coming. Because these network-connected devices are typically running Android, a shorter way of saying they are running Linux, self-replication becomes a very real possibility across the “internet of crap.”
Vendor Consolidation and Systemic Vulnerability: We saw the physical impact of connected infrastructure when cyberattacks targeted refrigeration systems in French supermarkets. Specifically, Internet-connected controllers with known critical vulnerabilities were the problem. As market consolidation limits competition to one or two incumbent vendors, hardware homogeneity means a single critical vulnerability becomes systemic, affecting an entire industry overnight. This risk extends to other connected hardware as well, as seen in reports on security cameras, alongside zero-click spyware campaigns, and Microsoft increasing overall patch load by +10x.
Corporate Espionage and Insider Threats: Looking past nation-states to industrial espionage, lawsuits are already emerging around team members moving from Company A to Competitor Company B and allegedly taking proprietary assets with them. Recently, honey pots and insider threats have been making headlines.
Frontier Lab Sandboxing and Regulatory Scrutiny: The boundaries of AI testing sandboxes are being pushed. AI models are “paying” with wallets they do not own. Similarly, nations are reporting AI agents are reading confidential materials undetected until AI model providers noticed and disclosed the incidents. These events will likely prompt governments and regions to demand clear explanations of how frontier labs handle sandboxing. We are already seeing movement toward both self-regulation and formal oversight: Anthropic is providing METR to third-party assessors, while the US FTC is demanding OpenAI and Anthropic provide more information and disclosures.

🏗️ AI Infrastructure

While AWS (with Graviton, Trainium, and Inferentia) and Google (with TPUs) have long established their custom silicon efficacy, Microsoft’s story around Maia and Cobalt has historically lagged, until now.

Microsoft’s Third-Generation Silicon: Microsoft announced Maia 300 and says multi-gigawatt deployments are starting next year.
Data Center Economics and the Green Energy Glut: The capital required for these deployments is substantial if you consider a 1GW AI data center could cost ~$30B (50% of the cost from the chips and the rest in power, construction, water, land, etc.), while a 500 MW data center might be ~$7B. When thinking about gigawatt-scale power, the challenge isn’t just generating renewable energy, but capturing and sequestering it so it isn’t wasted. For example, China extimates ~10% of its renewable output was wasted in 2026.
Fabs and Supply Chain Timelines: Every region is looking toward vertically integrated domestic chip manufacturing rather than assuming chips will come from halfway around the planet. GlobalFoundries is building in New York. Meanwhile, TSMC is building in Arizona, and Texas. Additionally, Taiwan is investing in Texas as the value chain. Even the earliest of these timelines are at least two years out, with larger projects taking three to five years, and the massive buildouts stretching a decade out.

💰 Funding Updates

Startups with an AI focus across healthcare, agentic workflows, security, and data center hardware continue to secure significant venture capital, with valuations climbing rapidly the moment a company touches core infrastructure.

Healthcare & Medical Devices:
- Suki: AI for healthcare providers, raised a $90M Series E.
- Scarlet: Certifies AI and software for medical devices, raised a $50M Series B.
Agentic Platforms, Collaboration & Enterprise Automation:
AI Security & Governance:
AI Infrastructure & Data Center Hardware:
Domain-Specific AI, Physical Systems & Fintech:

🤝 M&A Updates

Merger and acquisition activity continues to reflect companies acquiring capabilities that round out their enterprise portfolios, secure legacy code, or expand into physical AI.

Physical AI and Robotics: Omron will acquire OnLogic and Interlogic, for $675M to grow a “physical AI” portfolio. Meanwhile, SoftBank may buy Allied Universal for ~$18B.
Cybersecurity, Quantum, and Open-Source Remediation: Having covered IonQ several times before on the blog, it is notable that IonQ will buy Elisity for $1.3B, giving IonQ a clear foothold in cybersecurity. Additionally, OpenAI may buy Trail of Bits. If completed, this lets OpenAI tell enterprises that in addition to writing new software, they can help fix forgotten, unpatched open-source software before it is targeted by AI-powered cyberattacks. In the private equity realm, familiar territory from years of attending Qlik Connect, another Thoma Bravo portfolio company, Proofpoint may buy Varonis Systems for ~$6B.
Enterprise Workflows, RevOps, and Agent Identities: On the customer service front, Decagon may be acquired by ServiceNow. In financial software, Intuit acquired Sequence for ~$200M, expanding further into RevOps and billing systems. Finally, Workday will acquire C1.ai. Given what ServiceNow and Salesforce have been doing with their respective agent stories, Workday, ServiceNow, Salesforce, and Atlassian are going to be colliding more and more around managing human and agentic identities.

Thank you for listening to What The Fudge for Sunday, October 18, 2026. Be safe and be well.

#OpenToWork Weekly

I participated in a #NeverSearchAlone #JSC after rebooting Cuthrell Consulting and joining NexusTek.

I added those #OpenToWork from my LinkedIn network to this newsletter. Over time, we’ve grown an Alumni 🎉 section for those that transitioned to their next career adventure.

Alumni 🎉

Those in my network that are #OpenToWork:

View this page on GitHub.

What The Fudge for October 18, 2026: AI in the Workplace, Security Paradoxes, Funding, and M&A

Enterprise AI spend limits, IoT and frontier model security paradoxes, custom silicon and fab timelines, AI startup funding rounds, and strategic M&A.

Enjoyed this post?

Related Reading