IDC: AI Infrastructure Spend Hits $623B, Security Grows to $8B

Key Takeaways

• Cycode rebranded around agentic development security as AI adoption among developers hits 84%
• LinearB found heavy AI users merged 96% more PRs but agent PRs wait over five times longer for review
• Accenture survey: 72% of consumers expect AI to influence half their insurance shopping within 12 months
• IDC forecasts $623 billion in AI infrastructure spending, with security growing from $200 million to $8 billion in five years
• OpenAI launched Human Evaluation tool to test AI on everyday tasks like shopping and meal planning
• Cal Poly opened a $3 million AI Factory with NVIDIA support for student research
• Adipec 2026 revealed nearly 75% expect AI to increase energy consumption by 2030
• Moflin AI soft toys helped elderly residents become more social in German care homes
• Divyansh Agrawal compiled a RISC-V CPU into Qwen3.5-4B weights to run DOOM inside an AI model
• Cinch AI Marketing Teammate launched October 20, 2026, for auto service shop marketing campaigns

Cycode rebrands to focus on AI-era code security

Cycode rebuilt its brand around a new category called agentic development security. The company argues that AI has turned every employee into a developer and every attacker into a faster one. Gartner named Cycode a Leader in its June 2026 Magic Quadrant for Software Supply Chain Security. A Stack Overflow survey found 84% of developers use or plan to use AI tools, but Veracode research found leading AI models produced secure code only 56% of the time on average.

LinearB launches Agentic Studio for AI code context

LinearB launched Agentic Studio to give engineering teams plain language answers and callable context for coding agents. The agent skills are in early access, including an effort skill that routes work by risk and complexity. LinearB's 2026 benchmarks covering 2.7 million pull requests found heavy AI users merged 96% more PRs, but agent pull requests waited over five times longer for review. Token spend at the 90th percentile reached $35.20 per developer per coding day.

AI shopping recommendations risk losing customer trust

An Accenture survey of over 25,000 consumers in 16 countries found that 72% expect AI to influence at least half of their insurance shopping within 12 months. Researchers found 37% of loyal consumers would let an AI agent switch them to another brand for a better option. A Synchrony and Oxford Economics survey found three-quarters of consumers would shop less or stop using a brand entirely if its assistant recommended a poor product. Adobe research showed AI-driven traffic to retail websites jumped 62% year-over-year in July.

Cinch AI Marketing Teammate launches at SEMA 2026

Cinch AI Marketing Teammate became generally available on October 20, 2026, as an add-on to the Cinch platform. The tool lets auto service shop owners describe marketing goals in plain language and builds campaigns from customer data across email, text, mail and voice. Attendees at SEMA 2026 can see live demos at Cinch Booth 11532. Cinch is headquartered in Draper, Utah.

Adipec 2026 highlights energy's role in powering AI

The annual Adipec gathering in Abu Dhabi opened with AI as a central topic for the energy industry. Nearly three quarters of survey respondents expect AI to increase energy consumption by 2030, while only 52% believe existing infrastructure can meet the additional demand. The four-day event features discussions on modernizing grids, expanding storage and accelerating cleaner generation. Adnoc is advancing its oil production ecosystem to account for these developments.

AI soft toy Moflin helps lonely elderly in German care home

The Moflin is an AI-powered soft toy made by Casio that looks like a small furry animal. Ten Moflins were tested at a senior care facility in Hamburg, Germany, starting in late 2025. A study of 50 Moflins across five care homes found that over half of residents showed genuine interest and became more social. The toy provides warmth, soft breathing sounds, and noises to seem alive. Experts say it cannot replace human contact or pets but can help when those are not available.

AI could threaten capitalism by concentrating wealth in few hands

A Reddit post raised concerns about AI taking jobs and leaving wealth in the hands of only a few trillionaires. The post highlights fears that companies like Meta, OpenAI, and Amazon could own everything. The article compares this worry to ideas from economist Karl Marx about capitalism creating conditions for its own overthrow. The article says the future of capitalism is not looking bright but it is unclear if AI will cause a revolution or suppress one.

AI networking spending grows but enterprises are not ready

Spending on AI infrastructure is rising fast, driven mostly by hyperscalers and neocloud providers. Dell'Oro Group reports that enterprise ambition is outpacing readiness, especially in networking and storage. IDC forecasts $623 billion in spending, but network automation and skills are not keeping up. DDR4 memory shortages are pushing up prices for switches. AI system security spending is expected to grow from $200 million in 2025 to $8 billion in five years.

AI model runs DOOM using a RISC-V CPU compiled into its weights

AI engineer Divyansh Agrawal created the Qwenputer by compiling a 32-bit RISC-V CPU directly into Qwen3.5-4B weights. The CPU runs inside the AI model and can execute C code and play DOOM. The model uses Gated DeltaNet state as RAM to store instructions. Qwenputer has doubled the LLM size to almost 19GB, requiring a powerful workstation. The project is experimental and MIT-licensed, exploring what is possible rather than solving a specific problem.

OpenAI launches Human Evaluation tool to test real-world AI tasks

OpenAI announced a new tool called Human Evaluation to assess how AI handles everyday tasks. The tool tests AI on activities like shopping and meal planning. OpenAI says its models perform these tasks with high accuracy and sometimes outperform humans. However, the AI still struggles with understanding the nuances of human language and behavior. The company says this tool is a key step toward building more advanced AI systems that can assist people in daily life.

Cal Poly opens $3 million AI Factory with NVIDIA support

Cal Poly launched a $3 million AI Factory at its Swanson Center during the school's first AI Day. President Jeffrey Armstrong and NVIDIA executives attended the event that featured research displays and a digital activation countdown. The facility gives students and teachers access to advanced machine learning tools. Student researchers showed projects using AI in farming and business. People can use the factory after finishing a free training course through the Cal Poly portal.

Sources

• Cycode Rebrands Around a Blunt Thesis: In the AI Era, Security Must Start in the Code
• LinearB sells context as the fix for AI code
• AI is changing how consumers shop, but bad recommendations could cost brands
• Cinch AI Marketing Teammate Debuts at SEMA 2026 Showcase
• At Adipec, energy’s big chance to power the AI boom
• Can AI soft toys ease loneliness in older people?
• How AI Threatens Capitalism and Encourages Socialism
• AI networking spending grows, but enterprise network readiness lags
• AI model gets a 32-bit RISC-V CPU compiled directly into its weights, then runs DOOM on it — experimental AI model can run Doom and C code using its own weights as RAM
• OpenAI says its AI is getting better at doing things humans do, like shopping and meal planning
• Cal Poly unveils $3 million AI Factory to boost research