Observed 006

Hi there

Welcome to "Observed," your direct line into the lively discussions that shape our work at Obsolete. Now that the dust has settled we thought we’d take a look at the contradictions driving AI development. On one side, Deep Seek demonstrates powerful AI models running efficiently on standard hardware. On the other, the Stargate initiative proposes spending half a trillion dollars on AI infrastructure in Texas.

Small is Beautiful v’s Big Tech

Deep Seek has shown capabilities matching high-end commercial models while maintaining remarkably modest computing requirements. Major platforms like Perplexity AI are already integrating it as a replacement for more expensive options. This suggests a future where sophisticated AI capabilities become accessible to organisations of all sizes, running on local infrastructure without massive investment.

The implications are significant - businesses could deploy powerful AI models without excessive cloud computing costs or data privacy concerns. It's a vision of AI becoming a utility, available to all who need it.

Meanwhile, SoftBank, OpenAI and Oracle announced plans to invest $500 billion in AI infrastructure through the Stargate initiative. This massive investment suggests a future where AI capabilities are concentrated in hyperscale data centres, controlled by a handful of tech giants.

The timing and political context of this announcement - amid increasing US-China tensions and concerns about AI competition - raises questions about whether this is more about headlines than practical necessity.

Where's the Truth?

At Obsolete, we see foundation models trending toward commodity status. The real value isn't in raw computing power but in practical implementation and human expertise. Only 15% of UK SMEs have meaningfully engaged with AI - suggesting the challenge isn't computing capacity but practical knowledge and implementation capability.

The Stargate announcement feels like a continuation of the "move fast and break things" mentality that has dominated tech for decades. But as AI becomes more accessible and efficient, do we really need more massive data centres? Or do we need more investment in training, literacy and practical implementation?

A Third Way

The reality likely lies between these extremes. While some applications will always require significant computing power, many business needs can be met with more modest infrastructure. The key is matching capability to actual business requirements rather than chasing the latest headline-grabbing announcement.

Why It’s Important:

These contradictory developments highlight a crucial choice for organisations approaching AI adoption. The massive infrastructure investments grab headlines, but the trend toward more efficient, accessible models suggests a different path. Success will come not from raw computing power but from thoughtful implementation focused on solving real business problems.

For most UK organisations, particularly SMEs, this means focusing on practical implementation and building internal capability rather than worrying about infrastructure wars. The democratisation of AI capabilities creates opportunities for organisations willing to take a measured, strategic approach to adoption.

Next time someone mentions spending half a trillion on infrastructure, maybe ask why they're not investing in people instead.

That’s it for this week’s "Observed." As always, we’re committed to thoughtful and ethical exploration of AI’s potential. Stay tuned for more insights from our Show and Tell sessions as we continue to shape the future of AI together.

The Obsolete team