Big Ideas for 2025, Startup Valuations, thinking small, and why relationships matter

The risks are not as big as you perceive.

Good morning

In today's edition, among other things:

• Big Ideas for Tech 2025

• The Art and Science of Startup Valuation

• Why Relationships Are More Than Networking

• Two AI Futures

• Anatomy of High-Converting Homepages

• Thinking Small

• Cost of Failure and Competitiveness

Onwards!

---

Big Ideas for Tech 2025

Here’s an interesting summary of big ideas and tech from a16z that I think should be top of mind entering 2025:

AI Paradigm Shifts

• Limitations of synthetic data and pretraining: Current approaches are nearing their scalability limits.

• Shift in costs from training to inference: Real-time applications take precedence over upfront model training investments.

• Dramatic cost reduction in inference: Processing costs have dropped by 100-200x.

• Speed is critical: Users demand instant results, reshaping deployment priorities.

Impact on Public Markets

• Misaligned market focus: Investors overestimate NVIDIA's dominance while undervaluing companies further down the value chain.

• Distributed network architectures: Emphasis on low-latency, decentralized systems replaces large data clusters.

• Energy and network design assumptions: Updates needed for optimizing AI infrastructure for sustainability and revenue alignment.

• AI's growing market share: Now comprising 40-45% of total market capitalization.

Early-Stage Startups

• Small teams, big results: Teams of 2-5 people can achieve cutting-edge innovations.

• Lower capital requirements: Reduced funding needs empower more startups.

• Open-source models lead innovation: Meta’s Llama and similar models drive industry transformation.

• AI scarcity as a growth lever: Innovation driven by resource limitations, not demand.

• Shortened enterprise sales cycles: From months to minutes, enterprise AI adoption accelerates.

Key Industry Players

• OpenAI: Strong consumer brand faces competition from free alternatives by Meta and Google.

• Anthropic: Technological excellence but lacks distribution strategy.

• Google: Rich resources but unclear adaptation to the new paradigm.

• AWS (Amazon): Well-positioned as an infrastructure provider in the inference-centric world.

• Meta: Open-source strategy disrupts the competitive landscape.

• Microsoft: Strategic advantage through OpenAI ownership, but post-AGI dynamics are uncertain.

Broader Insights

• AGI may arrive by 2025: Rapid progress could lead to Artificial General Intelligence within two years.

• Underutilization of audio/video data: These remain less exploited compared to text-based AI applications.

• Excess enterprise GPU capacity: Over-purchases during the AI boom now create surplus.

• Acceptance of breakthroughs: Innovations like passing the Turing Test are becoming normalized.

Energy, Hardware, and Space

• AI drives nuclear energy revival: Reactivating plants and building new reactors to meet AI data center demands.

• Cross-disciplinary engineering demand: Engineers proficient in both hardware and software are increasingly critical.

• Space conquest accelerates: Reusable spacecraft like Starship enable new orbital infrastructure, from AI data centers to biomedical labs.

Biotech and Health

• Focus on common diseases: Success stories like GLP-1 inspire a return to tackling prevalent illnesses.

• Health tech democratization: Wearables and AI empower patients with actionable health data.

• AI in healthcare staffing: Automating administrative tasks alleviates labor shortages.

Consumer Applications

• AI-specialized video generation: Tailored solutions for specific use cases improve video quality.

• AI memory banks: Personalized digital archives analyzed for deeper insights.

• AI for knowledge work: Customized tools adapt to individual workflows and writing styles.

• Unstructured data analytics: Bridging qualitative insights with quantitative analysis.

Cryptocurrency

• AI-managed crypto wallets: Autonomous systems handle portfolios and assets.

• Decentralized AI chatbots: Independent entities perform complex functions.

• Identity verification: New systems tackle authentication challenges in the AI-driven world.

• Crypto app distribution: Emerging platforms reshape application ecosystems.

Enterprise and Fintech

• AI-driven compliance: Domain-specific language models streamline regulatory processes.

• Service industry transformation: Enhanced scalability through AI.

• Native AI interfaces: Reinvented UI/UX paradigms cater to AI-centric software.

Gaming Innovations

• AI-powered storytelling: Interactive Pixar-like experiences redefine entertainment.

• AI companions: Engaging virtual personalities create new social dynamics.

• Anonymous creators: AI enables content creation without revealing identity.

Growth-Stage Technologies

• Search disruption: Google’s dominance challenged by AI chatbots.

• Sales renaissance: AI-driven automation revitalizes sales professions.

Infrastructure and Development

• AI data center competition: Nations vie for leadership in building high-capacity AI infrastructure.

• Edge AI dominance: Localized models emerge as critical for numerous applications.

• AI reasoning advancements: Progress continues in complex fields like mathematics and physics.

It’s a lot, but think about it: it’s not everything. A very interesting year ahead.

---

The Art and Science of Startup Valuation

There’s a primer on startup valuations from EY. At its simplest, valuation is the process of determining what a company is worth. For founders, it’s not just a number—it’s a reflection of the company’s potential and a tool to negotiate equity stakes with investors. A higher valuation means retaining more control while raising funds. For investors, it’s about balancing the risk-reward equation. Overpay, and the investment may never generate returns; underpay, and you risk alienating founders or missing out on competitive opportunities. It’s art. It’s science. It’s both, and it matters (I’ve seen firsthand how some founders get valuation very wrong and very right).

Valuations arent’ “steady-state” but more “chaotic systems of potential”. The Discounted Cash Flow (DCF) method is most often used as a framework because it quantifies future cash flows, but its rigid structure clashes with the dynamic, uncertain nature of startups. Let’s dive deeper.

The DCF method evaluates a company’s value based on its projected future cash flows, discounted back to its present value. This discounting accounts for the time value of money (future cash flows are worth less today) and the inherent risks of realizing those projections. It’s a popular method because it prioritizes future performance over historical data—ideal for startups where future potential is the primary asset.

Consider the example:

Valuation is a philosophical exercise at some point. The DCF method anchors itself in two fundamental truths:

• Future Value Matters More Than Past Performance: Startups, often pre-revenue, are bets on potential rather than established returns. The DCF’s emphasis on projected cash flows aligns perfectly with this future-forward lens.

• Time-Value of Money: Money today is worth more than money tomorrow. This intrinsic discounting of future cash flows underscores the uncertainty baked into every startup bet.

But here’s the problem with the DCF method: startups thrive on uncertainty, while DCF demands predictive precision. This clash is the biggest issue—it assumes a deterministic future for entities defined by stochastic variability. The method has some more downsides to consider:

• Over-Reliance on Assumptions: The DCF’s output is only as credible as its inputs. For startups, these inputs—market size, growth rates, and cost structures—are often speculative.

• Sensitivity to WACC and Growth Rates: Small changes in these parameters can yield wildly divergent valuations, underscoring the fragility of the exercise.

• Blind Spot for Intangibles: DCF struggles to quantify factors like founder charisma, network effects, or first-mover advantage—elements that often determine a startup’s trajectory.

I think the output-to-input problem is the biggest one. The pace of change, number of experiments, stage you are in, product-market-fit, go-to-market phase - it’s difficult to get the inputs right.

But here’s my take on this: if someone is using DCF to evaluate your company - fine as long as you don’t have to do it. At the end of the day, from a founders perspective the final valuation figure emerges not from formula but from negotiation. Spend more time on being great at the latter.

---

Why Relationships Are More Than Networking

Building professional relationships isn’t about the size of your LinkedIn connections or the frequency of your emails. It’s about: