Build, Blend, or Rent?

Executive Summary 

In this edition, we explore the rise of humanoid robots as Norwegian startup 1X prepares to test them in homes, dive into Nvidia's groundbreaking AI tools and record-breaking revenue, and spotlight Synthesia, the leader in AI video creation, now valued at $2.1 billion.

Plus, our discussion post examines the varying strategies companies use to integrate AI into their ecosystems.

News

 🤖Humanoid robots are coming soon. Link

• Originating from Norway, robotics startup 1X plans to test its own humanoid robots in several hundred homes in 2025.

• 1X raised $125 million from notable investors such as Tiger Global and OpenAI, among others.

• The robots use AI to walk and navigate. While this initially appears impressive, humanoids are not capable of fully autonomous operations and rely on human teleoperators for actual activities.

 🔮Nvidia’s plan for the future. Link

• Nvidia introduced Isaac GR00T, a new tool to help create robots that can perform human-like tasks, especially for factories and automation.

• Nvidia launched a powerful platform called Blackwell Ultra AI Factory to help companies use AI more effectively and at a larger scale.

• Company achieved record revenue of $39.3 billion in Q4 fiscal 2025, marking a 78% increase year-over-year.

AI tool of the week

Synthesia

Synthesia creates professional tutorial videos in minutes. Simply input your text script, and Synthesia transforms it into a polished video.

This UK-based startup recently raised $180 million in a Series D funding round led by NEA, bringing its total valuation to $2.1 billion. Synthesia now serves over 60,000 customers globally, solidifying its position as the leader in enterprise AI video communications.

However, Synthesia has its limitations: its avatars can feel emotionally detached, making them less effective for deeply engaging content. The platform also struggles with long videos, occasionally producing jerky animations.

Discussion

Build, Blend, or Rent?

There are varying degrees of confidence in the future of AI technology. On one spectrum, people believe that AI models will be commoditised, with everyone having access for free or at a very low cost. Others believe that only a few market players will be able to develop and maintain top-of-the-market products, so language-model-as-a-service (LMaaS) is the way to go.

This split is mirrored in corporate strategies. Some rely on fully internal AI infrastructure, while others prefer to rent it.

Which approach is more productive?

While we don’t have insights from secret board meetings, we can analyse their AI vision based on the executed strategies. AI product companies are generally divided into three cohorts:

Owners

• These companies invest heavily in building fully integrated AI infrastructures. They aim for complete control over their AI capabilities.

• Companies such as Google, xAI, and Meta have fully integrated proprietary AI stacks. Mark Zuckerberg outlined a goal to invest $65 billion in the hardware ecosystem this year, while Google’s investments are targeting $75 billion this year alone.

• This approach enables easier and faster integration with their own universe of digital products.

• As a drawback, this strategy requires constant participation in the AI arms race to justify previous spending and maintain a competitive edge.

Hybrids

• Hybrid companies strike a balance between developing their own models and integrating external solutions for specialised tasks.

• Apple, a prime example, has developed its own Apple Foundation Models (AFM), which operate both on-device and through servers. These models are specifically optimised for iOS and macOS functionality and are more compact than ChatGPT's models, resulting in lower training and operational costs.

• Apple collaborates with OpenAI for more advanced and processing-heavy use cases (e.g., image processing and advanced conversations with Siri).

• This approach allows Apple to prioritise security and maintain a conservative approach in terms of its own infrastructure spend.

Renters

• Renters rely entirely on external providers for their AI needs, focusing on integrating these solutions into their existing products.

• Salesforce and Perplexity are prime examples in this category. Marc Benioff clearly outlined the policy that Salesforce will leverage the infrastructure of others without risky investments in its own language models (LMs).

• This strategy enables companies to provide multiple LM options to end users while focusing on the development of their core product offerings.

Conclusion

We can see that companies have different approaches to AI ecosystems, and, looking at the current players, it is difficult to conclude if any specific strategy is beneficial.

Having analysed the market, I believe that companies are trying to leverage their product-market fit in the most efficient manner. If securing their market position requires building their own model, they build; otherwise, they rent.

Our service

Do you need help with your data ecosystem? Reply directly to this email or reach out to [email protected]. More information about our service is available here.