Technology FAQ

The answer depends on which metrics you prioritize. a16z points to 80% GPU utilization and $40.6 billion in annual revenue from top AI companies as evidence of real demand. AQR counters that the U.S. …

The CAPE ratio measures price relative to 10-year average inflation-adjusted earnings. At current levels around 40x, it suggests U.S. large cap equities may return roughly 3.9% annually over the next …

Yes, this is one way to reconcile the bullish and bearish views. Individual AI winners generating real revenue and high utilization could deliver spectacular returns even as broad market indices …

Both sources have financial incentives influencing their perspectives. a16z manages billions in venture capital and growth equity, so bullish AI narratives support their portfolio valuations. AQR runs …

Tech P/E multiples currently sit around 30-35x, elevated but nowhere near the 70-80x of 2000. More importantly, GPU utilization runs at 80% compared to just 7% for fiber optic cables during the dotcom …

A hybrid recommender system combines multiple recommendation approaches, typically using fast, cheap methods like collaborative filtering for candidate generation from millions of items, then invoking …

Netflix uses a three-layer architecture: offline systems train deep collaborative filtering models on viewing history, nearline systems update user embeddings seconds after interactions, and online …

The exploration-exploitation tradeoff is the balance between recommending items users will likely enjoy (exploitation) versus surfacing new content they might not discover otherwise (exploration). …

LLM-based recommendations cost orders of magnitude more than classical methods. A single LLM recommendation can consume thousands of tokens, while traditional collaborative filtering uses simple dot …

Spotify's AI DJ uses an "agentic router" that decides whether to invoke expensive LLM reasoning or fall back to fast keyword matching against collaborative filtering embeddings. Complex queries get …

AQR's 2026 capital market assumptions show U.S. buyouts returning 4.2% versus 3.9% for public equities. That's a 30 basis point premium for accepting years of lockup, unpredictable capital calls, and …

According to AQR's 2026 capital market assumptions, expected real returns for U.S. buyouts are 4.2% over the next 5-10 years. Private credit is even lower at 2.6%, down 0.5 percentage points …

Manager selection is critical. In venture capital, top decile managers generate 31.7% IRR while bottom decile managers return negative 7%. The spread between winners and losers is enormous. But this …

Too much capital chased the same opportunities. When every pension fund, endowment, and sovereign wealth fund allocates 20-30% to alternatives, the returns that made them attractive get arbitraged …

Access to companies you can't reach in public markets. 87% of U.S. companies with more than $100 million in revenue are now private. Value creation has shifted earlier: for 2020-2023 IPOs, 55% of …

Evidence suggests yes. Sara Hooker's research shows that compact models like Llama-3 8B now outperform massive predecessors like Falcon 180B. Scaling laws only reliably predict pre-training test loss, …

Yes. Llama-3 8B outperforms Falcon 180B, and Aya 23 8B beats BLOOM 176B despite having only 4.5% of the parameters. This is not an isolated fluke but a systematic trend visible in Open LLM Leaderboard …

Hedge fund short interest in utilities now sits at the 99th percentile relative to the past five years, according to Goldman Sachs data. The bet appears to be that AI data center demand may not …

As scaling hits diminishing returns, the skills that matter shift toward algorithmic cleverness, data quality, and architectural innovation rather than brute-force compute. Frontier labs are …

Yes, in the cryptographic sense. Statistical analysis of 20,000 rounds confirms the random number generator produces fair outcomes matching the stated 97% RTP. However, mathematical fairness doesn't …