The Strange Tariff Calculus: Stocks have risen further just because tariff levels aren’t as ridiculously high as first announced, but still sit at levels not seen since the 1930s. Disrupted supply chains need to be rebuilt, and either inflation eats into the spending power of Americans or company margins shrink if they do not pass on tariffs. Neither of these environments lays the groundwork for higher equity valuations. Eventually, rational thinking should return and reevaluate stocks at levels that are more sensible.
Soundbite: The OBBB’s AI tax breaks and the release of America’s AI Action Plan added fuel to already frothy AI sentiment. Last week’s deregulations will streamline the permitting process for building data centers and accelerate semiconductor manufacturing. The tax bill frees up cash flow for the hyperscalers by allowing for the immediate and full expensing of domestic R&D costs, along with deductions for capex on servers and networks. Federal support to ease the way for companies to build major infrastructure projects and promote AI exports was welcomed, but we cannot help but recall Ronald Reagan’s words: “The nine most terrifying words in the English language are I’m from the government and I’m here to help."
Last Wednesday’s plan contained major policy shifts. The plan featured lower obstacles to building and powering data centers, and steps to boost exports of AI tools. Specifically, the changes include:
· The federal government will impose fewer regulations, such as environmental concerns, on the construction of new data centers and energy infrastructure.
· Make Federal land available for data centers and power generation facilities.
· Advance AI innovation and adoption, such as open-source and open-weight AI models.
· Expanded Federal agency export support for U.S. AI firms and relaxed restrictions on shipping their products abroad, including banned chips exports to China.
· Training programs for workers needed to staff facilities, and retraining for displaced workers.
· Ensuring access to large compute for startups and academia by promoting spot and forward financial markets for computing power.
· Withholding aid to states with regulations that could slow technological innovation and the effectiveness of any federal funding program.
Brave New World New Deal: the effort is expansive and impressive, and a deep public-private partnership of this scale holds promise but also potential pitfalls. We are concerned that the “move fast and break things" mentality does not translate to the federal government.
All firms benefit from the tax-cutting AI provisions contained in the OBBB, but it will widen the gap between the Magnificent-7 and the rest of the S&P 500 companies. Jeffries estimates that the new R&D tax rules and capex deductions will increase free cash flow (but not earnings) of nonfinancial S&P companies by “6.7% and 8.5% respectively, in 2025 and 2026.” However, given the enormous capital spend of the Magnificent-7, hyperscalers are the major winner with an expected cash flow boost of over 12% for this year alone.
The stars seem to have aligned for an even bigger moonshot for mega cap tech when we add in the bullish demand from corporate stock buybacks that kicks in over the next few weeks. Yet, we cannot forget that the VIX index currently hitting lows also aligns perfectly with a seasonal low in volatility that rises sharply into October. We are left to wonder if all the booster rockets have been fired and are already priced into stock prices.
Soundbite: There is a well-known adage in Silicon Valley that people overestimate a technological breakthrough in the short term but underestimate it in the long-term. We are certainly in the overestimation stage. Limitless hyperscalers’ capex binge plus the administration’s action plan to achieve global AI dominance has raised hopes for tangible productivity improvements. There are expectations of a regime shift from long-standing 2% trend growth to a “welcome to the new age” 20% annual GDP growth. That assumption translates to a one-year AI-powered productivity boost that equals twice the entire productivity increase from the Internet. While we cannot pinpoint a time, this bubble almost certainly pops when these forecasts are lowered, even if they are revisited a few years later.
A 10-fold jump from established productivity growth requires a re-write of the theory that marginal increases in capital push productivity lower. Growth has always been a function of labor force growth plus productivity. Therefore, Epoch AI forecasting 20% GDP growth surprised us. Assuming AI will eliminate jobs, yielding negative labor force growth, you need productivity growth in excess of 20% to meet Epoch AI’s future GDP forecast. That means productivity increases accelerate over time, instead of gradually decaying as is normally the case. We believe that is a stretch, at least in the short term.
The reality is that just to maintain baseline productivity growth of 2% requires a technological breakthrough—no breakthrough causes below-trend productivity. While we believe an even better outcome than the internet’s enormous impact is a reasonable assumption, it is a leap to embrace such lofty GDP estimates.
America enjoyed 1.8% productivity growth for 20 years before the internet, then it jumped to over 3.0% from 1999 through 2005, thanks to the pervasive acceptance of the technology. Since then, with no new major advances, the U.S. economy has reverted to 1.6% productivity growth. The seven years when the internet drove productivity 1.3% above the previous two decades generated a cumulative real growth total of roughly 9% (1.3% annual increase multiplied by seven years).
Take note of that statistic: total real excess growth of 9% is attributable to the internet, an incredible achievement. To put these current growth forecasts into perspective, the promise of AI is to more than double that figure…each year.
Perception is all that matters for stocks. Perhaps this time around the sun, things will be different, and investors will not make the same mistake from 1999-2000 about what breakthrough technology can deliver in the near term. Productivity has a tight relationship with real earnings growth (and real GDP growth), so it does not surprise us that NVIDIA’s 25% and 30% respective compounded earnings and revenue forecasts match the sci-fi sized 20-30% GDP forecasts.
One warning: Just because numbers match assumptions does not mean they are rational.
Therefore, the hype is immediate, but the rollout is not: when overheated perceptions collide with reality, markets become hectic. They did in 2000, and at some point, the two will meet again.
Soundbite: Thinking about AI is a game of extremes. It is difficult not to view the future through asymptotic eyes when looking at AI-produced goods that can trend toward a zero price tag. Alternatively, it is easy to envision the exponential possibilities of infinite returns on capital or skyrocketing demand for money pushing up interest rates astronomically. We are wondering if the best forecasts are right down the middle: energy constraints, errors in human judgement, institutional friction toward adoption, or superintelligence that gets writer’s block. Any of those possibilities can slow down superlative growth to a more humanistic scale.
This week’s Economist magazine’s cover story is on the “Economics of Superintelligence”, and it has many interesting thoughts. They pinpointed what we had written above: if you add machines (meaning capital) but not labor, capital becomes less productive unless the technology itself can deliver new technological improvements. The underlying belief in what we consider unreachable GDP estimates is that almost fully autonomous AI labs will conduct their own scientific research within a few years.
Our main issue is that for growth to increase by an order of magnitude above its current trend for a decade, ideas must beget more ideas at an accelerating rate. Therefore, progress must be multiplicative. Before that happens, which tech futurists see as inevitable, we expect to encounter massive bumps in the road. Our inevitable scenario contains intolerable budget problems if the superintelligence scenario materializes due to the ever-increasing expected demand to fund AI projects. Massive debt held by China, the U.S., Japan, and the European Union sovereigns will have to deal with substantial increases in real interest rates that compound budget deficits to painful extremes.
Hopefully, artificial general intelligence will find a solution to that problem. While we like to establish balance in our Three Pointer when markets lean too far one way or the other, we are optimistic at our core. The world will find its way, even if it is not the path that anyone expects.
(All times E.S.T.)
1. Role Reversal: the main event is normally either the Unemployment report or the Fed Meeting, but this week the biggest event could be the combined releases on Monday, July 29, of the Treasury Borrowing Advisory Committee report and Wednesday, July 30 of the Quarterly Refunding Announcement. The key items will be what is the highest acceptable threshold of Bill Issuance of the Federal government relative to the total auction amount, as defined by the Advisory Committee and then the actual issuance portion of short-term Treasury securities. J.P. Morgan expects a 23% share as the highest acceptable level for Bills to be auctioned.
2. Wednesday, July 30 at 2:00 p.m. Federal Open Market Committee meeting rate announcement and statement. Chair Powell’s press conference starts 30 minutes afterwards. We will see if Governor Waller dissents to the likely unchanged rate policy and whether he is joined by Governor Bowman. The press conference may give color on whether September or October will be the possible timing of the rate cut.
3. Employment Trifecta: Friday, August 1 at 8:30 a.m. Nonfarm Payrolls report for July. The focus will be on the split between Government and Private payrolls. June showed Government employment was over half the overall gain at +74,000, while it is only forecast to be +5,000 in June. Consensus is for Private payrolls to be just over +100,000 with an uptick to a 4.2% unemployment rate. Thursday, July at 8:30, Initial Claims, which has fallen for 5 straight weeks. Released at the same time is Continuing Claims (still near its highs) and the Employment Cost Index for Q2. Tuesday, July 29 is the June JOLTS data, and we are looking, as always at the Hirings and Quits rate. Normally Thursday’s 8:30 a.m. broadcast of June’s Personal Consumption Expenditure Price Index would be ignored after Wednesday’s FOMC rate announcement, but if it comes in soft, the calls for 300 basis point immediate rate cuts will be coming out of the White House at 8:31 a.m.
FOMC Voters Speaking: No Fed speakers scheduled this week after the Wednesday rate announcement. The Bank of Canada’s Interest Rate Decision is out a few hours before the Fed on Wednesday, July 30 at 9:45 a.m. and is also expected to show no change in policy followed by their 10:30 a.m. press conference. The BoJ rate decision is announced at 11:00 p.m. Monday, July 28.
Earnings: Mega Cap Tech AI Spend Focus as over one-third of the S&P 500 reports this week: Wednesday, July 30, after market close for Microsoft, Meta, QUALCOM and Lam Research. HSBC and UBS report before market open. Thursday, July 31 after market close for Apple and Amazon, with Mastercard reporting before market open. Tuesday, July 29 after market close is Visa, giving us the first look for the week from a major player discussing consumer spending.