Peter Garnry, CFA

Peter Garnry, CFA

@petergarnry.gesdacapital.com

CEO & Co-founder at Gesda Capital. I’m passionated about equity investing and equity valuation. My vision is to show that active equity investing can work if set up properly. I was previously the Chief Investment Strategist at Saxo Bank.

625 followers193 following829 posts

I find it mind boggling that so few investors are talking or reflecting on this. One this is circular vendor financing, but outright cross ownership made casually between a growing group of US technology companies is bad for competition over time and amplifies financial risks one day.

Peter Garnry, CFA@petergarnry.gesdacapital.com

The US Keiretsu continues.

The US Keiretsu continues.

Jesse Felder@jessefelder.com

‘Nvidia, Advanced Micro Devices and Broadcom have agreed to buy as much as $5 billion of the common shares in aggregate. Meta intends to use the proceeds to support its investment in AI infrastructure and personal superintelligence.’ www.bloomberg.com/news/article...

Unfortunately quite predictable and something I have written about more than a year ago. When you increase CapEx margin will go down in the future unless OpEx is cut. These companies are transforming into capital intensive low employee to revenue companies.

Tom Hearden@followtheh.bsky.social

*MICROSOFT TO OFFER BUYOUTS TO 60,000 WORKERS IN LATEST COST CUT 27% of its workforce!!!!!

Anthropic's S-1 filing has been filed. 1H figures show that the company is not a long way from being positive on operating income. And there are not talking about the ridiculous EBITDA figure but the real operating income.

View full thread

selling compute right now at spot given the supply constraints. I infer that SpaceX's AI figures are reflecting that selling compute spot to Anthropic is an okay business at current prices. If they fall over time the economics will shift from infrastructure compute to foundational and model layer

Anthropic's S-1 filing has been filed. 1H figures show that the company is not a long way from being positive on operating income. And there are not talking about the ridiculous EBITDA figure but the real operating income.

It seems all the cash they have raised is not really necessary as CapEx intensity is low and they are almost profitable. So the cash is predominately for potential acquisitions. The figures also suggest when combined with SpaceX, that the foundational layer is less profitable, than

I guess it is not a big surprise. The domains where LLMs make a big difference are not people intense relative to the economy. I remain of the view that the productivity gains will come from physical AI and research breakthroughs. Hassabis’ direction is a good indication of this.

Dare Obasanjo@carnage4life.bsky.social

Apollo’s chief economist finds that there is still no measurable impact of AI adoption on corporate America outside of big tech. Revenue per employee and profit margins are rising for the Magnificent Seven but flat for the other 493 S&P 500 companies.

The fact that profit margin is only going up in technology sector is a sign of circular revenue and business flow among n companies in the technology sector. Productivity (revenue per worker) looks depressing and especially if you adjust for inflation.

Mike Zaccardi, CFA, CMT@mikezaccardi.bsky.social

Is the AI boom a bubble outside Tech? 📉 Rev per worker: S&P 493 flat 📉 Profit margins: S&P 493 flat 📉 Productivity boost: None Non-tech sees AI token spend as a cost center, not growth. The macro AI story will be decided in the S&P 493, not the Mag 7. T at Apollo

Reflecting that investors are realising ROIC will decline as the depreciation wave hits and CapEx slows leading to lower revenue growth and lower operating margins.

Mike Zaccardi, CFA, CMT@mikezaccardi.bsky.social

The forward 12-month P/E ratio for the S&P 500 is 18.7. This P/E ratio is below the 5-year average (19.8) and below the 10-year average (19.1) @factset

It all started with the Google IPO allowing poor governance to fester. Because Google was so successful, Silicon Valley could sell the benevolent unconstrained founder story to Wall Street ever since.

Scott Horton@robertscotthorton.bsky.social

The Meta board consists almost entirely of Zuckerberg's cronies and sycophants, and now he treats them exactly the way Musk treats his board—essentially as a rubber stamp for whatever he wants. The need for trust busters and heavy enforcement against abusive plutocrats is here.

SpaceX figures are giving us clues to the ROIC figures on AI. In this case Grok. Looking at the annualised change in NOPAT over the past year for the AI segment relative to the estimated CapEx invested over the period, the incremental ROIC looks to be around 12.6%.

This was using gross figures. Using net figures which is more correct the incremental ROIC increases to around 15% excluding the Cursor acquisition’s impact on incremental invested capital which would take incremental ROIC down to half the level.

SpaceX figures are giving us clues to the ROIC figures on AI. In this case Grok. Looking at the annualised change in NOPAT over the past year for the AI segment relative to the estimated CapEx invested over the period, the incremental ROIC looks to be around 12.6%.

This is a respectable figure but not impressive. Given the risks that SpaceX is taking, I would not assume SpaceX's cost of capital is less than 10%. Probably much closer to the incremental ROIC. Things can change for the better, but it is first look into the profitability at the model layer.

Depreciations are also coming with a lag, so both the incremental ROIC and long-term ROIC will fluctuate quite a bit over the coming years. And soon Cursor will begin to influence figures, so the future ROIC analysis of the AI model layer will be a bit less pure vs Grok only.

When everyone speeds up with AI, differentiation will come from going deliberately slower. Excessive development speed and change due to AI could become like a volatility drag on companies lowering your long term growth. Counter intuitive to the wisdom of Silicon Valley.

One of the more important charts on equities right now. Given the depreciation lag incremental ROIC is much lower and given they are all not going to be winners one of them will potentially destroy a lot of value.

Mike Zaccardi, CFA, CMT@mikezaccardi.bsky.social

T at Apollo: Return on invested capital for the five largest hyperscalers is declining, as invested capital grows faster than the earnings it generates

Really important point and key risk to understand. Although this was also one of the key risks to social media spending but as long as the VC ecosystem has positive momentum it can continue for a long time…

Ara Kharazian@arakharazian.bsky.social

This chart shows the key risk to the AI trade, and it's missing from the latest revenue announcements. The AI boom is fundamentally driven by a top 1% of companies that drive 80% of spend. Within that top 1%, 80% are venture-backed.

Red flag 🚩 on the board of directors at Ralph Lauren. This incentive structure is obviously wrong. It’s like setting up a bonus structure only on new sales without taking gross margin into account. Guess what will happen?

Dare Obasanjo@carnage4life.bsky.social

Ralph Lauren replaced sustainability targets in executive bonuses with AI adoption targets. The four AI KPIs were 2 AI training completion rates and 2 employee AI survey response rates. The board approved a +10% bonus modifier for exceeding them. None measured actual productivity gains from AI. 🤷🏾‍♂️

I participated in the CFA Institute survey on Trump’s idea through the SEC to remove mandatory quarterly reporting for public listed companies to semiannual reporting. The results are in. 62% support retaining mandatory quarterly reporting. 70% oppose broad issuer flexibility on frequency.

Had an interesting experience at the hospital today. I was there to get a blood sample and while I was sitting in the room there was this small conveyor belt running behind me. Suddenly others blood samples were on the belt moving past me and in the direction into another room.

I asked the nurse whether the conveyor was taking the samples to the analysis room and he replied yes. He said it was so smart. Previously they spent quite a bit of time walking back and forth. We constantly talk about AI and its productivity gains.

But this was a small real life productivity example of someone coming up with a simple idea to improve a tedious work process. It was a funny experience that made me smile. It was literally like running sushi in the most unexpected place…

Essentially the AI deflation wave Paul Kedrosky has been talking about all year…

Jesse Felder@jessefelder.com

‘This slowdown has been unusually stubborn. We’ve recorded two straight months of virtually no growth in token spend. The main driver is price declines at the frontier, in addition to efficiency and quality improvements for models across intelligence levels.‘ econlab.substack.com/p/ai-index-o...

Very happy user…has never for a second regretted leaving Chrome behind.

Vivaldi Browser@vivaldi.com

ok Bluesky people, here´s a challenge for you: 4.5M people have switched to this Norwegian AI-free browser and said "enough is enough" to Big Tech and its user-tracking AI-down-everyone's-throats-malpractices. Can we make it to 5M before 2026 is over? 😌 🙏

Agree that it distorts incentives but ads are a necessary “tax” for those that want things for “free”.

Dare Obasanjo@carnage4life.bsky.social

OpenAI plans to show ads while generating images. As someone who works in ads, I dislike the incentive this creates. If ad views depend on image generation time, creating images faster could reduce ad revenue. Ads making the experience so bad that you’re compelled to pay reminds me of YouTube.

Peter Garnry, CFA reposted

Annual revisions lifted the savings rate level, but the sharp downward trajectory remains. Despite all the noise about AI, the biggest driver of strong US growth over the last 2.5yrs has come from household's spending growing much faster than their incomes.

The rumours on Nubank being in early talks about acquiring UK-based Monzo took down the stock price by 10% yesterday. Investors are fearing shareholder dilution, lack of focus and too high premium. On the other hand, Monzo is fast growing a very successful digital bank.

Monzo is planning a broader European launch but it will require more tier 1 capital and more scale. I can see the synergies but as a shareholder I have the same worries as the market. But we have high confidence in David Velez and the management team so for now we believe in their decisions.

Incredible downfall for Nike. The most outrageous thing is that it could much more given the bleak sales outlook. The ultimate test of brand strength…

Carl Quintanilla@carlquintanilla.bsky.social

B of A throws in the towel on $NKE: “.. We are cutting F27E/F28E EPS by 11%/12%; we now expect negative sales growth through F27E versus our prior view of a Spring inflection.

Det danske herrelandshold i fodbold har en massiv ledelseskrise. Det har stået klart længe og ansættelsen af Kent Nielsen var et klart tegn. Du må ikke havne et sted hvor en anfører ventilerer sine følelser i pressen og basalt stiller et ultimatum mod holdets bedste spiller.

En ting er at det sender et signal om en ledelseskultur i ubalance. Men det sender også dønninger ind i truppen og kan forvolde ganske meget skade. Riemers interview efter kampen med David Nielsen og Bo Henriksen udstiller også den manglende autoritet.

Det er fuldstændig misforstået, hvis man tror, at en landstræner skal gå ind i et studie og hyggesnakke med kommentatorerne samt acceptere alle kommentarer på spillet. Pilen peger på Peter Møller og DBU.

Peter Garnry, CFA reposted

Search interest in “Muse” was still higher in September 2009 than it is now (That’s when the band Muse released its amazing “Resistance” album)

The AI token deflation wave is the largest ever since in the modern economy. Some will survive this but the ROIC economics will be tough to manage. It will require an insane operational cost focus. Completely opposite modus operandi of the past for hyperscalers.

Antoine Walker Capital@antoinewalkercap.bsky.social

H/t @nuancerocket.bsky.social

Google search is dying. The question is how long it will take. And more importantly, can Google offset search revenue with YouTube, cloud and AI?

Jesse Felder@jessefelder.com

‘Traffic to news sites is plummeting in the U.S. as tech giants shift users from search and social sites to AI interfaces.’ www.axios.com/media-trends...

Meta’s Muse is interesting but is it a most. That is the market’s current verdict but Amazon’s rejection says that the path is complicated. Interesting point sequence: Claude Code - software Mythos - cybersecurity Muse - banks The first narrative has often been wrong. Could happen again…

I love when investors talk about this era is exponential across so many dimensions. Sorry to break the party. The economy has always been exponential. What they actually mean is super exponential but the physical world will destroy that dream before it even starts…

Peter Garnry, CFA reposted

From looking at data from OpenRouter, Ramp & Vercel, we can see the AI business model in real time. Anthropic remains the monetization leader, but revenue growth is slowing as customers shift to cheaper models. OpenAI is going the other direction, with usage and spend on its models accelerating.

Of course this is the strategy. Any rational person should see this and not believe this extinction nonsense. The paradox is that Trump flips the narrative on the call with Jensen saying it a hoax. My take is that it is all about a giant political battle shaping up…

Dare Obasanjo@carnage4life.bsky.social

It’s a fair point that Dario’s AI slowdown proposal’s toughest constraints aren’t on American frontier labs. They’re on China. The strategy is to restrict chips, compute and distillation to widen the US lead, then use that lead to give American labs room to slow down.

Extremely valid point and likely something that could be extended to other domains in our new geopolitical era.

Phillips OBrien@phillipspobrien.bsky.social

Latest piece. People asked why the Saudi military seems hamstrung? A problem it and other Gulf states face is that they have constructed their militaries on systems that they buy and do not make. phillipspobrien.substack.com/p/why-buying...

Peter Garnry, CFA reposted

Last time the Fed hiked before an election? 2022 midterm under Biden. From 1984 to 2024, the Fed adjusted rates (up or down) within 60 days of an election (43%) as often as in any other 60-day period (45%).

Aaron Rupar@atrupar.com

Hassett threatens Warsh: "I'd be wary of a rake hike. The Fed has had hiked rates ahead of an election only a couple times all the way back to 1913. So I think if you want an independent Fed, one thing the Fed does is stay out of the way of elections."

Management says that despite the change in RPO to 664bn… “there is no incremental impact on its plans to raise capital”. This due to customer prepayments which amounted to ~11bn in the quarter. This is where Nvidia comes into the picture. They help with the financing through circulation of capital

Another issue is that their supposedly massive increase in operating income does not reflect the true underlying profitability of the new assets. With the massive deflation in token prices and competition from open weight models I don’t see how this is sustainable. Time will tell…