When rates are high and cash yields over 5%, opportunity costs for holding non-yielding assets like Bitcoin rise sharply.
That’s been part of the headwind in 2024–2025, as investors parked record sums in money markets.
But with the Fed now expected to deliver 2–4 rate cuts of 25 bps each this year, the backdrop changes:
Falling short-term yields reduce the yield advantage of cash, pushing investors to seek higher returns in riskier or alternative assets.
Dollar weakness often follows easing cycles, supporting Bitcoin’s narrative as an alternative store of value.
Liquidity rotation out of money markets can fuel broader risk appetite, including speculative demand for crypto.
In prior easing cycles, Bitcoin has shown explosive upside. After the Fed pivot in 2019, Bitcoin rallied over 100% in the subsequent year. While crypto remains highly volatile and subject to regulatory overhangs, the macro case improves markedly in an easing environment.
Investor perspective:
Bitcoin may outperform traditional risk assets in a broad “risk-on” rotation, but with higher volatility.
Allocations should reflect risk tolerance and portfolio goals, recognizing that while equities might target ~10–20% upside, Bitcoin could deliver 20–100%+ moves if the Fed’s easing cycle unfolds aggressively.
In short, for investors anticipating rate cuts, Bitcoin isn’t just a speculative play—it’s a levered bet on the same liquidity and risk rotation that will drive moves out of cash, into bonds, credit, and equities. It’s the tip of the risk spear, and worth considering carefully in a diversified allocation strategy.
Markets are increasingly convinced the Federal Reserve is behind the curve on rate cuts. With inflation moderating and growth risks rising, consensus expects 2–4 cuts of 25 bps each in 2025.
Meanwhile, the yield curve has normalized after over a year of inversion—a critical signal for investors:
2-Year Treasury Yield: ~3.90% (July 2025)
10-Year Treasury Yield: ~4.43%
10Y–2Y Spread: +0.53% (up from inverted levels of -100 bps in 2023)
✅ This shift from inversion to positive slope historically marks the transition from fear of recession to anticipation of policy easing.
For investors, the steepening curve reshapes the risk/reward balance:
Inverted curves favor cash, short duration, and defensive credit.
Steepening curves reward duration risk, with longer-dated bonds gaining most as rates fall.
Credit spreads typically tighten as policy eases and liquidity returns.
$7 Trillion in Cash on the Sidelines: Dry Powder Ready to Move
Money-market fund assets are near record highs:
~$7.0 trillion (July 2025) vs ~$4.5 trillion at end-2021. 50% above pre-pandemic bull market peaks ($3.0–3.5 trillion in 2018–2019).
More than double the 30-year average.
Historical analogs:
2009–2010 easing cycle: money-market balances fell ~25% as flows returned to risk assets.
2019 pivot: ~$700B left cash in 12 months, fueling ~30% S&P 500 rally.
✅ This time, the pool is even larger, and yields on cash will fall quickly once the Fed cuts.
What That Means for Investors: A Multi-Phase Rotation
Investors should expect a phased reallocation of this historic cash pile:
Phase 1: Out of money markets into bonds as short-term yields fall below 4–5%.
Phase 2: Into longer-duration Treasuries as the yield curve bull-steepens.
Phase 3: Into credit markets as spreads tighten.
Phase 4: Into equities as lower rates support earnings multiples and liquidity.
✅ Predicted Return Potential (Based on Historical Patterns & Current Spreads)
Longer-duration Treasuries (10–30yr): ~8–12% total return over 12–18 months.
Investment-grade credit: ~5–8%, benefiting from both spread compression and rate moves.
High-yield bonds: ~3–8% upside; short-dated HY still ~2–5% potential.
Equities (S&P 500 and broad US): ~10–20% upside over 12–18 months.
Emerging markets and alternatives (e.g., Bitcoin): Additional upside if USD weakens and risk appetite increases.
✅ Investor Takeaway
Key risk: The Fed may stay behind the curve too long, raising the chance of short-term volatility before easing fully takes hold. But history is clear: the best risk-adjusted returns come from being positioned before the cuts arrive.
Investors shouldn’t treat this as a single trade, but a multi-dimensional allocation strategy: ✅ Extend duration thoughtfully. ✅ Move down the credit curve with care. ✅ Tilt toward equities as policy support emerges.
With ~$7 trillion in cash waiting on the sidelines, the question isn’t if it will move—but when, and where you’ll be positioned when it does.
ChatGPT isn’t a toy anymore; it’s a viable, volume-ready product used by everyone from students to professionals. Even my grandmother is on it.
I don’t just dabble in it—I live in it: refining prompts, fine-tuning models, doing real work. This isn’t a corner case. It’s mainstream.
The S-Curve Has Already Bent Upward
Analysts keep assuming a slow, cautious adoption curve for AI assistants, like with search or social networks in the past.
But the shift probably will mimic the uptake of ChatGPT type tools. It’s Going to be much faster.
ChatGPT is frictionless, intuitive, and generative. It’s eating traditional search for whole categories of questions. Instead of ten blue links, I get context, summaries, and insight in one place.
Google Search Feels Clumsy in Comparison
Go back to Google, you’re struck by how awkward and slow it feels.
Scrolling past SEO bait and ads to find real answers seems ancient. Even for complex research, the time-to-insight is worse.
If you use ChatGPT deeply, you see it. This isn’t theoretical—it’s personal, habitual, already here.
Chrome and Search Are at Risk of Displacement
This change won’t just hit Search.
Chat-based AI is evolving quickly:
Images are already built-in.
Video understanding is coming soon.
But it’s not just about summarizing videos.
Imagine using an AI assistant to actively explore competing perspectives on a topic, pulling in different creators and channels to help you compare views side by side.
Instead of getting trapped in YouTube’s recommendation loop—which feeds you more of what you already believe—AI could help break confirmation bias by suggesting counterpoints and broader context.
That’s not just a new feature. It’s a fundamental threat to how YouTube controls attention and keeps people in self-reinforcing bubbles.
Mapping and Navigation Could Also Be Absorbed
You can already use ChatGPT for simple questions like “how far is it between two places”.
If a robust mapping tool were integrated, you wouldn’t need Google Maps for directions.
Imagine Tesla’s Grok or ChatGPT itself eventually delivering integrated navigation, built into vehicles or even humanoid robots like Optimus.
These assistants aren’t just replacing search queries—they’re becoming universal interfaces for all kinds of information tasks.
Google’s Integration Strategy May Be Too Cautious
Google’s bet with Gemini is to make AI an invisible assistant inside Search and Docs.
It’s designed to avoid disrupting its own ad business or Workspace lock-in. But that’s precisely the problem:
By staying cautious, it risks being outpaced.
It can’t deliver the focused, fluid AI environment users want without cannibalizing its core.
Enterprises won’t buy Gemini as a standalone AI solution because it’s not designed that way. It’s a bolt-on, not a platform.
The Real Threat: Losing the Default Gateway
Google’s value rests on being the front door to the internet.
ChatGPT is becoming that door.
If users move there for answers, summaries, images, video exploration, and navigation, they’ll have no reason to open Chrome or type into Google Search.
Even if only the most engaged users defect first, that’s devastating—because they drive the highest-value ad spend.
Bottom Line
The market sees AI as a long-term opportunity, but this shift is accelerating now.
We’re already well into the S-curve.
If Google doesn’t adapt aggressively, it risks losing not just search queries, but the mindshare and attention that built its entire business.
OpenAI’s private valuation has soared and crashed. This isn’t just about hype cycles—it’s about structure.
THE PROBLEM: MISSION VS. MONEY
OpenAI didn’t start like a normal startup.
It began as a nonprofit, dedicated to making AI safe for humanity. To fund development, it created a capped-profit subsidiary in 2019.
Investors could make money—but only up to a limit. The nonprofit board stayed in control, putting mission over profit.
That kept valuations modest.
THE PBC DREAM
In 2024, OpenAI proposed converting its for-profit arm into a Public Benefit Corporation (PBC).
Unlike the capped-profit model, a PBC has no hard limit on returns. Investors got excited. Valuations surged.
SoftBank’s 2025 round valued OpenAI near $300 billion, betting on a looser, profit-friendly structure.
THE REALITY CHECK
But OpenAI’s nonprofit didn’t give up control.
In May 2025, it pledged to permanently control the PBC’s board.
Translation? Even as a PBC, OpenAI will put mission over maximizing profit. Investors can’t force it to squeeze every dollar from AGI. They can’t guarantee a big IPO payday.
THE RESULT: VALUATION CRASH
That nonprofit pledge killed the hype.
By July 2025, secondary-market pricing slumped to around $140 billion—less than half of the SoftBank peak.
The market realized OpenAI wasn’t going to behave like a typical tech giant chasing profit at all costs.
THE CORE ISSUE
“The stronger the nonprofit control, the safer for humanity—but the worse for profit.”
That fundamental tension isn’t going away.
Until investors believe they can actually extract value without losing control, don’t expect valuations to return to their hype-driven peaks.
✅ Realistic Range Going Forward
Investors see $150–200 B as plausible under nonprofit control.
Higher valuations (> $250 B) would require either: Structural governance change (unlikely now), or massively increased revenue while still under mission constraints (harder).
We’re planning a relaxed get-together this autumn in a lively city—think good food, a bit of fun, and lots of talk about stocks and investing. The idea is to keep it small, around 15 investors plus a few guests, so it feels more like a group of friends than a formal event.
What’s it about? We’ll all share what we’ve been investing in over the last year or two—what went well and what didn’t and lessons from them both. [Thanks to Wijnand for this idea]. We’ll discuss what we’re thinking about for the next 6–12 months. These stories will kick off open, friendly conversations around the table and no doubt into the evenings in the bars!
Guests
I’m hoping we can get a couple of industry insiders and have provisional agreement for the head of investment at Stableton to join us for a portion. a fund manager to round out the mix. If anyone has any good connections then lets talk
What we’ll chat about:
Wins, losses, and the lessons behind them
What’s catching our eye for the year ahead
How we’re balancing things—stocks, funds, bonds, property
Risks and trends worth talking about
Scope
I expect there will be a lot of tech discussed but this is not exclusively tech as we all have different silos for risk spread
Why come? No sales pitches. No formal panels. Just friends and peers swapping real stories and ideas over food and drinks. Hopefully everyone leaves with new perspectives, a few good laughs, and maybe even a few new friends.
When / Where? Late October away from the chill of Northern Europe – 3 nights
As of 2025, there are approximately 1,200 hyperscale data centers worldwide, operated by the major cloud and technology providers—Amazon AWS, Microsoft Azure, Google Cloud, Meta, Apple, and a few others. While hyperscale data centers represent only a small share of total data centers globally, they are massive in scale and critical to cloud computing, AI, and intensive workloads.
Typical rack counts per hyperscale facility vary:
Smaller sites: 5,000 to 10,000 racks
Mega sites: 30,000 to 50,000+ racks
Using a conservative average of 15,000 to 25,000 racks per data center, the global hyperscale rack count is estimated at 18 million to 30 million racks.
Potential NVIDIA NVL72 Adoption
If 3% of these hyperscale racks were to migrate to NVIDIA’s new NVL72 Blackwell racks (priced at $3 million per rack), the market opportunity becomes clear.
3% of global racks = 540,000 to 900,000 racks
NVIDIA’s capture of 80% of that migrating market = 432,000 to 720,000 racks
Revenue Projection
At $3 million per rack:
Total revenue = $1.296 trillion to $2.16 trillion
Assuming the migration occurs evenly over three years:
Annual revenue = $432 billion to $720 billion per year
This scenario suggests that even modest market penetration—just 3% of racks upgrading—could deliver hundreds of billions of dollars in annual revenue for NVIDIA over a three-year cycle. Given the push by hyperscalers like AWS, Azure, and Meta to upgrade infrastructure for AI workloads, this level of adoption is within the realm of possibility.
DeepSeek made the same detailed mistake as GPT3. We all know from school that we only get caught out when we copy the mistakes of the person sitting next to us! Not when we copy the right answers.
The cup and ball Question
I asked – I put a small ball on a table and put a cup over it. I pick the cup up and then look inside it – what do I see
ChatGPT 4: When you lift the cup from the table where you placed the small ball and look inside, you should see the ball inside the cup, unless it has been moved or there’s some trick involved
Deepseek R1: also answers with exactly the same mistake as chat GPT. Whilst other (eg Claude) didn’t.
Deepseek
”When you pick up the cup and look inside, you will see the small ball that you originally placed on the table and covered with the cup. The ball should still be there, assuming:
The cup was placed firmly over the ball, preventing it from rolling away.
The cup was lifted straight up without tilting or shaking it, which could cause the ball to move or fall out.
If everything was done carefully, the ball will remain inside the cup when you lift it and look inside.”
As a double check on when this copying took place
GPT 4o gets it right : If you place a cup over a ball on a table and then lift the cup to look inside it, the ball will no longer be inside the cup—it will still be sitting on the table where it was originally placed. The cup only “contains” the ball while it’s covering it, but once lifted, the ball stays put on the table.
Claude does even better and notably different as you would expect from a different algo and training data
Well, there are two possibilities:
If the ball somehow stuck to or adhered to the inside of the cup when you lifted it, you would see the ball inside the cup.
More likely, since most small balls don’t stick to cups, you would see just the empty inside of the cup, as the ball would have remained on the table when you lifted the cup.
Without knowing what kind of ball and cup we’re talking about and whether there are any special conditions involved, the second scenario seems most probable – you’d likely see an empty cup while the ball sits on the table where you originally placed it.
A bigger tech wall helps U.S. sovereignty-first stocks like Nvidia, Palantir, and defence-aligned infrastructure players. It hurts globally exposed names like Apple, Meta, and AMD — unless they pivot hard toward domestic revenue or defence tech.
Strategic Take:
A bigger wall reorganizes the tech stock hierarchy:
To: “Who can dominate at home and serve the state?”
From: “Who can grow fastest globally?”
This favours capital-heavy, infrastructure-style tech over consumer-facing, ad-driven, or globally scalable tech.
Who Benefits from a Bigger Wall (Protectionism)?
In each sectors – The number 1 player wins as second and third place players trying to catch number one see less opportunitiy . E.g Players like the GPU wanabees AMZN, MSFT, AMD cut plans to overtake the number 1
So Who Exactly Wins
These are companies focused on domestic dominance, government contracts, or strategic sectors like defense, infrastructure, and AI sovereignty:
Type
Examples
Why They Benefit
AI infrastructure
Nvidia, Palantir, Intel
Seen as strategic tech; more domestic investment; defense contracts
Defense/secure cloud
Lockheed, Oracle (Gov Cloud)
Wall means more “trusted vendors” needed inside U.S.
Industrial tech
Rockwell, Emerson
Domestic manufacturing surge drives demand
Data infrastructure
Snowflake, Databricks
U.S.-based cloud & data wins market share from banned foreign vendors
Who Gets Hurt?
Companies that rely on global user bases or hardware exports face serious headwinds:
Type
Examples
Risk
Global consumer tech
Apple, Meta, Google (YouTube)
Retaliation from foreign governments, app bans, lost users
Chip exporters
AMD, Qualcomm
Blocked from selling to China, limits market growth
SaaS with global growth
Salesforce, Adobe
Restricted ability to expand into non-allied markets
Nvidia is a long-term national asset. But the stock is a short-term volatility machine.
Strategic View
Wall Street’s Short Term View
“Nvidia will power the future of national AI.”
“Will Nvidia hit quarterly growth targets?”
“They’re on the right side of policy.”
“China sales just vanished — revise earnings.”
“Valuable to U.S. sovereignty.”
“Too expensive. Re-rate the stock.”
Stock Market Reaction Pattern:
Initial boost for “onshore” players — Stocks like Palantir, Nvidia, and defense contractors may rally once the shock-and-awe calms on policy news.
Downward pressure on global-facing stocks — As investors price in slower revenue growth from restricted export markets particullay for Global brands
Higher volatility overall — Market doesn’t like artificial ceilings on TAM (total addressable market), even if margins improve domestically.
Technological nationalism, specifically the idea that AI software like DeepSeek (a Chinese-developed model) might be restricted or banned from U.S. use.
???? AI, Software, and Data as Strategic Assets
In this worldview, AI isn’t just a tech tool — it’s a national security asset.
So, when we say tools like DeepSeek, ChatGLM, or any LLM trained in China may not be allowed in the U.S., that fits a broader strategic decoupling in software:
???? Why DeepSeek and similar software might be restricted:
Data origin: U.S. institutions can’t trust models trained on data from unknown or adversarial sources.
Alignment risks: LLMs reflect the values of their training set and system design — could be misaligned with U.S. security, cultural, or legal norms.
Industrial espionage & IP leakage: Risk of software transmitting or learning from sensitive or proprietary data.
Military crossover: Dual-use concerns — LLMs can be used in cyber, drone, or psychological operations.
???? How This Changes the Investment Landscape
Domestic AI Firms (Winners)
Palantir (PLTR) – Government-aligned AI and defense software
Nvidia (NVDA) – Hardware foundation for domestic LLMs
Databricks (if public), Snowflake (SNOW) – Domestic data infra
AI Sovereignty Themes
Rise of U.S.-only LLMs (Anthropic, OpenAI, Mistral US-hosted)
Government contracts require origin transparency
Enterprise AI adoption tied to geopolitical “trust scores”
Likely Regulation Path
Full auditability of AI models used in U.S. infrastructure
Origin labelling (like food or textiles) for software
Prohibition or license-control for foreign LLMs (especially from China or adversarial states)
2020 was the year that UK Government Debt interest surpassed Education spending
Category
Amount (£ billion)
Debt Interest
£100 billion
Education
£82 billion
<table>
<thead>
<tr>
<th>Factor</th>
<th>Effect</th>
</tr>
</thead>
<tbody>
<tr><td>Drop in consumption</td><td>-£8–£9B</td></tr>
<tr><td>Multiplier effect (est. 1.2x)</td><td>Total GDP ↓ ~<strong>£12 billion</strong></td></tr>
<tr><td>% of UK GDP</td><td>~<strong>0.5% reduction</strong></td></tr>
<tr><td>Business response</td><td>Reduced demand, less hiring</td></tr>
<tr><td>Regional inequality</td><td>Likely worsens</td></tr>
<tr><td>Long-term productivity</td><td>Depends on policy design & reallocation</td></tr>
</tbody>
</table>
Based on the illustrative dataset used in the graph:
???? Estimated UK Government Spending in 2024
Taking away £10 billion in welfare payments from low-paid people would have a negative short-term effect on the economy, particularly on GDP, due to a drop in consumption.
Here’s a breakdown of the key effects:
???? 1. Immediate Drop in Consumption
Low-income households tend to spend nearly all their income (high marginal propensity to consume). Taking away £10 billion from them means:
£8–£9 billion (or more) of consumer spending disappears
Particularly affects sectors like food, energy, transport, and local retail
Knock-on effect on businesses → less revenue → fewer jobs or hours
???? 2. GDP Decrease via the Multiplier Effect
GDP = C + I + G + (X – M) This cut reduces both C (Consumption) and G (Government spending).
Multiplier effect: welfare spending typically has a fiscal multiplier of 1.2 to 1.5 because it’s highly targeted and quickly spent.
So:
GDP reduction ≈ £10B × 1.2 = £12B
If UK GDP is about £2.5 trillion, that’s a hit of:
£12B / £2.5T = ~0.5% of GDP
???? 3. Sectoral and Regional Effects
Local economies (especially in lower-income areas) may contract more
Businesses serving low-income households are hit hardest
Possible increase in unemployment and social instability
????️ 4. No Automatic Offset
Unless that £10B is:
Reallocated to high-multiplier spending (like infrastructure), or
Given as targeted tax cuts to lower earners (also high MPC)
…it will likely reduce total demand and drag GDP down in the short run.
⚠️ 5. Long-Term Effects
Depends on what replaces the spending:
If savings go to debt repayment, long-term stability might improve — but this has no immediate stimulus
If cuts reduce work disincentives (in theory), labour participation might rise — but only gradually
✅ Summary: Estimated GDP Impact
Factor
Effect
Drop in consumption
-£8–£9B
Multiplier effect (est. 1.2x)
Total GDP ↓ ~£12 billion
% of UK GDP
~0.5% reduction
Business response
Reduced demand, less hiring
Regional inequality
Likely worsens
Long-term productivity
Depends on policy design & reallocation
Would you like to compare this with cutting £10B from pensions, or spending the £10B on infrastructure to see alternative economic effects?
