Future Bytes

For this week’s AI News: Three orders of magnitude now separate US, Chinese, and European AI bets. We unpack the week's $700B numbers and the four questions every business leader should bring to their next planning session.

What Stood Out This Week:

• US AI infrastructure spend hit a new scale. The four big hyperscalers reported $130B in combined capex for one quarter, with full-year 2026 tracking above $700B.
• Anthropic raised $65B in five days, then got shut out of the Pentagon. Classified AI contracts went to Amazon, Google, Microsoft, Nvidia, OpenAI, SpaceX, and Reflection.
• China's AI stack is now complete: Own models, own silicon, own buyers.

Hosted on Acast. See acast.com/privacy for more information.

For this weeks AI News: Four major AI models launched in five days.

GPT-5.5 went agentic, DeepSeek hit the frontier at one-tenth the price, and Anthropic bet $100 billion on compute. The era of standardizing on one model is over.

Top stories for week 18:

• GPT-5.5 launches as an autonomous agent built to complete tasks, not assist with them
• DeepSeek V4 Pro hits frontier quality at one-tenth the price, on Chinese silicon
• Anthropic signs a $100 billion AWS compute deal and resets API defaults to Opus 4.7
• Google launches Gemini Enterprise, a single runtime for 200+ models including Claude

Hosted on Acast. See acast.com/privacy for more information.

PwC just quantified the AI divide: 20% of companies are capturing 74% of the value. Within days, it showed up on a stock chart as Claude Design launched and Figma fell 12%.

Top stories for week 17:

• PwC finds 20% of companies now capture 74% of AI's economic value
• Claude Design launch sends Figma down 12%, deepening the SaaS apocalypse
• Opus 4.7 ships while Mythos stays restricted
• LeCun vs. Amodei clash reframes the jobs debate around re-skilling

Hosted on Acast. See acast.com/privacy for more information.

Hosted on Acast. See acast.com/privacy for more information.

An AI model capable of uncovering thousands of unknown vulnerabilities was built and not released. The decision, and the response from governments, signals a shift: AI is no longer just accelerating innovation, it’s exposing risk at scale.

Top stories for week 16:

• Anthropic withholds a frontier model after discovering large-scale zero-day vulnerabilities
• Thousands of security flaws identified across widely used systems, most still unpatched
• Governments convene emergency meetings to assess systemic AI-driven cybersecurity risks
• Cybersecurity moves to the top of the agenda as AI changes the threat landscape

Hosted on Acast. See acast.com/privacy for more information.

One of the world’s most safety-focused AI companies exposed its own source code due to a simple configuration error. The incident points to a broader shift: as AI moves into production, the real challenge is no longer the models themselves, but the governance, processes, and operational maturity around them.  

Top stories for week 15: 

• Anthropic incidents highlight how operational risk can impact even leading AI players 
• Most enterprises are still early in scaling AI agents, with limited visibility and control 
• New solutions emerge to manage agent identity, access, and governance 
• Enterprise platforms accelerate deployment of AI agents across core business workflows 

Hosted on Acast. See acast.com/privacy for more information.

Hosted on Acast. See acast.com/privacy for more information.

We are heading into Easter break with a new episode of Future Bytes News, your weekly updates on AI in business. This time, host, Magnus Oxenwaldt, VP Group AI at Columbus covers these top stories:  

OpenAI abandons a billion-dollar Disney deal to focus on enterprise.

Anthropic advances with AI coworkers and cybersecurity-grade models.

The shift signals that AI is moving from consumer hype to business-critical infrastructure. 

Hosted on Acast. See acast.com/privacy for more information.

AI is moving beyond models and into workflows. In this episode, host Magnus Oxenwaldt, take a closer look at OpenClaw’s rapid rise and what it signals about the shift toward agent platforms and orchestration layers. We also explore the gap to enterprise deployment, and what it means for your business and AI strategy.

Highlights from week 13:

·     OpenClaw becomes the most starred project in GitHub history, signaling rapid adoption of AI agents.

·     Value shifts from models to agent layers that connect AI to tools and workflows.

·     Security risks expose the gap between experimental agents and enterprise readiness.

·     Big tech shifts strategy toward platforms, workflows, and AI infrastructure.

Hosted on Acast. See acast.com/privacy for more information.

Four companies launched AI coworkers in a single week, all designed to sit alongside the world’s knowledge workers. At the same time, new data from OpenAI co-founder Andrej Karpathy maps which roles are most exposed. Is this where the shift begins? 

Hosted by Magnus Oxenwaldt, VP Group AI at Columbus. 

Highlights week 12: 

• Microsoft, Google, Anthropic and Perplexity launched AI coworkers built for knowledge workers. 
• These systems execute tasks across tools, moving from assistance to workflow automation. 
• New data shows 42% of jobs are highly exposed, especially high-paid, screen-based roles. 
• Enterprise software is being repriced as AI reduces the need for multiple tools and licences. 
• AI vendor choice is becoming a strategic and geopolitical consideration. 

Hosted on Acast. See acast.com/privacy for more information.