Newsletter · · Ashutosh Agarwal

Alphabet and Tesla's AI Capex Bill Comes Due as Oil Hits $100 - US Macro Recap - Week of July 24, 2026

Global macro newsletter for the week of July 24, 2026. Alphabet and Tesla both raised AI capital spending and posted negative free cash flow for the first time in years, oil spiked back toward $100 a barrel on renewed Iran and Red Sea escalation, and Wall Street forecasters split sharply on whether next week's Fed meeting brings a hike, a hold, or the start of cuts.

US Macro Recap

Week of July 24, 2026: Alphabet and Tesla's AI Capex Bill Comes Due as Oil Hits $100

For months the podcasts framed the AI building boom as the thing holding the whole economy up. This week the boom sent the market a bill, and the market flinched. Alphabet and Tesla reported, cranked their spending plans higher, watched their cash flow turn negative, and had their worst day in more than a year. At the same time, oil ripped back to $100 a barrel on renewed fighting around Iran, dragging bond yields up with it. And all of this lands four days before the Federal Reserve meets on Tuesday, with the professionals who forecast for a living now openly split on whether the next move is a rate hike or nothing at all.

TL;DR

  • The AI spending story stopped being abstract. Alphabet raised its 2026 capital-spending plan to $195–205 billion (from $180–190 billion) and posted its first negative free cash flow in about two decades; Tesla more than doubled its own spending to $5.8 billion and also went cash-flow negative. Both stocks had their worst day in over a year. The comforting idea that computing power is so scarce it justifies unlimited spending is now being openly questioned, and the bond market is starting to act like a disciplinarian.
  • Oil back at $100 reintroduces the inflation problem at the worst possible moment. Just as one camp was declaring the inflation fight over, Brent crude jumped back to roughly $100 on renewed Iran and Red Sea escalation. That reset the whole debate heading into Tuesday's Fed meeting and next week's PCE inflation report. The sell-side is now genuinely divided: PGIM put out a note calling for three rate hikes, while Goldman Sachs and Morgan Stanley both expect the Fed to sit on its hands.
  • The "K-shaped consumer is collapsing" story got its first serious pushback. A widely shared note argued the narrative is "roughly half reality and half overstatement," pointing to bank data showing card delinquencies at their lowest since September 2023 and lower-income wage growth re-accelerating. But credit-data firms still flag rising stress among the most vulnerable borrowers and a typical household that has actually gotten a bit poorer since 2021. Both can be true.

What's new

1. The AI building boom sent the market a bill, and got punished for it. (Operator/insider voices: company CFOs on their earnings calls, plus the CNBC trading desk.) This was the defining event of the week. On Squawk on the Street (July 23), the hosts played Alphabet's CFO from the prior night's call: "We are updating our full-year 2026 CapEx guidance range to $195 to $205 billion, up from our previous estimate of $180 to $190 billion… we continue to expect our CapEx to increase significantly in 2027." The kicker: the company posted its first negative free cash flow (spending more cash than it brought in) in roughly two decades, and did no buybacks for a second straight quarter because, as the hosts put it, "they can't: they're issuing" debt. Tesla, on the same call cycle, "more than doubles" its capital spending "to $5.8 billion," also producing negative free cash flow, with Elon Musk insisting "all the things that we're investing in will yield incredible returns… the best CapEx returns that we've ever seen." Both stocks had their worst day in more than a year. Nvidia's Jensen Huang, in an Axios interview clip aired on the show, waved it all off: "in the future, every single industry will be CapEx heavier." Why it matters: last week the podcasts described AI spending as "a credit story." This week that story became concrete and negative: the companies raised spending, went cash-flow negative, and the market sold them anyway. As one host warned, "the bond market is going to make them disciplined at a certain point."

2. The Fed forecast splits wide open: one desk calls three hikes, two say sit still. (Insider voices: sell-side and buy-side economists.) Heading into Tuesday's meeting, the professional forecasters cracked into two camps reading the same data to opposite conclusions. On Bloomberg Surveillance (July 21), PGIM's Rob Sockin defended what the hosts called "a shockingly terse note on 3 rate increases": "all the fundamentals to us point toward upside risk to inflation… above-trend GDP growth… a labor market we think is heating up… supply-side challenges: energy, but also in trade, labor." His central worry is non-shelter services inflation: "the majority of core PCE… running at high levels, shown no progress over the last year", and he reads the whole Fed as reaching "a hawkish critical mass," with Chair Kevin Warsh focused on getting inflation "back to 2.0% without rounding." The other side is just as credentialed. On Goldman Sachs' Exchanges (July 21), chief US economist David Mericle called last week's soft June inflation print "a bit of an outlier" but still the start of "a run of softer inflation data," arguing that once you strip out tariffs, "the CPI is running close to 2% on a core basis." His base case: Fed on hold this year, next move a cut in 2027, and he pegged the odds of a hike at "more like 25%," versus market pricing he reads as "a hike and a half." Morgan Stanley's Michael Gapen, on Thoughts on the Market (July 21), agreed inflation "has already peaked," sees it falling "to about 3% by the end of this year and closer to 2.5% next year," and expects no move, while admitting the risk is that "inflation just lingering at elevated levels could mean the Fed gets brought in to raise rates in September or later this year." Why it matters: this isn't a pundit-versus-pundit shouting match. It's three of the most serious research shops on the Street looking at identical numbers and landing on "three hikes" versus "on hold", which is exactly why Tuesday's meeting and next week's PCE report matter so much.

3. Oil back at $100 puts the inflation layer right back on the table. (Insider voices: a macro-policy economist and a cross-asset CIO.) On Bloomberg Surveillance (July 23), Julia Coronado of MacroPolicy Perspectives laid out how the oil spike changes the math: "$100 Brent… certainly brings another round of inflation pressures on top of several layers that we've already gotten: from the earlier phases of the war, from the AI buildout, from tariffs. There's just layer upon layer now of upward pressure on inflation." Crucially, she argued the pressure is broadening beyond energy: core inflation is "rippling through airfares… through software prices… still through core goods prices," and the goods-price relief everyone expected in the second half "is now at risk because shipping costs are going to go up." Morgan Stanley's cross-asset CIO James Caron, on the same show, made the point that complicates the Fed's job: this is "a supply-side shock," and "monetary policy is generally there for demand-side shocks," so hiking into higher oil prices "doesn't really match up." His conclusion is to stay "underweight interest rate risk… underweight bonds" and broaden into "value… healthcare… and actually, the consumer." Why it matters: June's friendly inflation number was, in large part, a gasoline story. With Brent back at $100, that tailwind reverses right before the Fed meets, and it hands the hawks fresh ammunition even though the cause is a war, not an overheating economy.

4. The consumer debate finally got a real counter-argument. (Pundit voices with hard bank data on one side; credit-data insiders on the other.) The "K-shaped economy": shorthand for the top half thriving while the bottom half struggles: got its most forceful pushback of the year on The Compound and Friends (July 21), where the hosts walked through a note from Adam Parker arguing the story is "roughly half reality and half overstatement." His evidence: lower-income after-tax wage growth "went from 2.9% in May to 4.1% in June" in Bank of America's deposit data; "90-day card delinquencies are the lowest they have been since September 2023"; 37% of credit-card accounts were paid in full in the first quarter while minimum-only payments fell to 10%; and 73% of adults told a Fed survey they were "doing okay or living comfortably." The hosts' framing: "the bottom of the K… they're not deteriorating… It's substitution and price sensitivity, not demand destruction," and pointedly, "the companies that report that the K is okay are the banks, they literally have the data, the companies blaming the lower consumer are fast-food companies." The counterweight came on Market Pulse (July 23), from Equifax, where Moody's economist Justin Begley warned that "credit stress is rising among the most financially vulnerable subprime cohorts, and there is risk that performance deterioration could become systemic if the labor market weakens further," and Equifax's Ian Wright noted the gap underneath the averages: total consumer wealth has grown 20% since 2021 to $74 trillion, but the median household's wealth actually "went down by 1%," with the bottom 54% of households (those under $100,000) seeing their assets shrink "by about 17%, 18%" over four years. His memorable image: a man with "his head in the oven and his feet in the freezer" has a perfectly normal average body temperature. Why it matters: this is the single healthiest disagreement in macro right now. The bank data genuinely shows the bottom half is paying its bills; the credit-and-wealth data genuinely shows it's getting squeezed and falling behind the top. The tie-breaker, as almost everyone agreed, is the labor market: if jobs hold, the bottom holds.

5. AI is a huge market story, but a much smaller economy story than the narrative claims. (Insider/analyst voices: a bank economist and a systematic-macro researcher.) A quieter but important thread pushed back on the idea that AI spending is single-handedly holding up GDP. On Morgan Stanley's Thoughts on the Market (July 21), Michael Gapen explained that although hyperscaler capex is running at "3.5% or so of GDP," roughly "60% of that… goes to items like computers and peripherals… that have a very, very high import content," so the actual contribution to US growth is "around 40 basis points": real, but "we just have to remember it's fueling growth around the world, just not here." He directly rebutted the popular claim: "I have seen some estimates where people say, well, if it wasn't for AI CapEx, the U.S. economy wouldn't have grown at all. And that's clearly wrong." On Excess Returns (July 23), systematic researcher Aahan Menon went further, saying his daily GDP nowcast shows the economy growing at "a really steady and kind of feeble pace," with tech capex adding only "somewhere between 10 to 30 basis points" to a headline near 2.7%: "the economy is mostly just consumption… and AI is a very small part of that basket." Why it matters: if the real economy is being carried by the consumer, not by AI, then the AI capex wobble this week is more of a market-and-credit event than a growth event, for now. But Menon's sharper point is that the economy is "more dependent on the stock market than it's ever been before," because falling savings and rising stock wealth are what's funding consumer spending. That links the AI selloff back to Main Street through the wealth effect.

The debate

This week's podcasts supported three distinct camps. All three had real, substantive voices, so all three are laid out below: with the honest note that the fault lines shifted this week because of oil.

Camp 1: "Disinflation is winning; the Fed should stay on hold (and eventually cut)." (Sell-side economists.) This camp gained the most institutional weight this week. Goldman's David Mericle (above) sees inflation going "more sideways, stay closer to 3% than to 2%" but reads the underlying trend, stripped of tariffs and war, as "reasonably close to 2%," and keeps the Fed on hold with a cut in 2027. Morgan Stanley's Michael Gapen sees inflation already past its peak and heading to ~3% this year, ~2.5% next. The steel-man: the June inflation data actually moved toward target, the breadth of the softness was unusual, tariff effects are fading, and the economy itself "is not creating an inflation problem": the pressure is coming from one-off supply shocks the Fed can look through.

Camp 2: "Inflation is sticky and broadening; the next move is a hike." (A buy-side economist and macro strategists.) PGIM's Rob Sockin is the loudest version, with an explicit three-hike call built on above-trend growth, a tightening labor market, and non-shelter services inflation that has "shown no progress over the last year." Julia Coronado's "layer upon layer" of oil, AI, and tariff pressure, plus core inflation broadening into airfares and software, supports it; so does James Caron's "higher for longer" call. The steel-man: five-plus years above target, a core-services measure still creeping up, oil back at $100, and a Fed that got burned by "transitory" last time and is determined to hit 2% "without rounding." As the State Street strategist put it on Street Signals (July 23), even after last week's soft print, "the market pricing for rate hikes this year most likely gets the sign or the direction of travel right, even if the magnitude is still very much up for debate."

Camp 3: "Forget hikes; the real signal is demand destruction and a cooling consumer." (Pundits and independent analysts.) A smaller but vivid camp read the same soft June inflation number as evidence the consumer is running out of money , not that prices are merely calming. On Tom Bilyeu's Impact Theory (July 23), the hosts argued that with headline CPI falling "nearly 0.5% in June," core CPI declining "for the first time since May of 2020," and even services-excluding-shelter turning negative, "there is much, much more evidence for demand destruction than there is even a hint of evidence for second-order inflation", and pointed to TIPS break-even rates (a market gauge of expected inflation) "plummeting" since late May as confirmation the market is pricing inflation out. David Rosenberg made the fundamentals version on Squawk on the Street (July 22): the economy is down to "3 props for GDP growth this year": AI spending, a war-induced inventory build, and "this radical, epic drawdown in the US personal savings rate", while real consumer spending has cooled from a 3% trend a year ago to 2%, held up only by one-time tax breaks. He argued core inflation excluding shelter is "running exactly at 2.0%" and that the true driver of inflation, unit labor costs, is running just "0.5%," concluding the Fed "is still a prisoner of missing transitory." The steel-man: if the softness is demand weakness rather than victory over inflation, hiking into it would be a policy error, and the credit-card data (delinquencies at a 15-year high in Rosenberg's telling, spending shifting from debit to credit "out of necessity") suggests the bottom of the K is leaning on plastic to get by.

The tension worth naming for readers: Camp 1 and Camp 2 are arguing about how sticky inflation is , while Camp 3 is arguing it's the wrong question entirely: that the bigger risk is a consumer quietly tapping out. And this week, the arrival of $100 oil strengthened Camp 2's hand on the print even as Camp 3 insists the cause is weakness, not strength.

The trades in play

Instrument-level positioning was unusually rich this week, and it clustered around one idea: yields go higher, and bonds are no longer the easy hedge.

  • Underweight bonds / bearish duration (strategists and a cross-asset CIO, insider view). Morgan Stanley's James Caron said on Bloomberg Surveillance (July 23) that "we can't just rely on bonds as… the easy hedge" anymore, favoring "underweight interest rate risk" and rotating toward value, healthcare, and the consumer, with the 10-year yield around 4.69%. On Power Lunch (July 22), Baird's Chris Verone and Schwab's Kevin Gordon were "bearish bonds… the trend in yields is higher," but flagged the reassuring part: "credit conditions… haven't deteriorated" and inflation break-evens are "relatively benign," so higher yields aren't yet foreshadowing a downturn. Rick Santelli noted the market is nibbling toward the 10-year's May high near 4.67% and that the move higher in rates is "global in nature": France's yields hit a 15-year high on the week.
  • The most contrarian macro call: Treasury yields to 10% (hedge fund manager, operator/positioning). On Monetary Matters with Jack Farley (July 22), London-based hedge fund manager Russell Clark argued the real bubble isn't AI, it's the Treasury market, and reiterated his year-end target of "Treasuries at 10% yield." His logic: to fix an unaffordable-housing problem you need wages rising "about 7% a year," housing flat in nominal terms, and a real rate near 3% "so people keep their money on deposit rather than sticking to real assets": a world of structurally higher inflation and rates driven by governments that "spend what we need to spend and won't tax anyone either."
  • Leaning back into gold with a cheap hedge (options trader, operator). On Macro Voices (July 23), Patrick Ceresna laid out his trade of the week: own gold (via the GLD fund) around $376 after its roughly 30% correction, wrapped in a "risk corridor" through September 18: buying downside protection from $370 down to $350 and capping the upside at $415: for a total cost of just $1.75 a share. The thesis, echoing guest Luke Gromen: this week the pattern flipped to "war on, rates up, oil up, gold up," different from the prior five months, suggesting the market is beginning to price war spending and "financial repression" as bullish for gold.
  • Currencies: fade the dollar near-term, own emerging-market carry (State Street strategist, operator/insider). On Street Signals (July 23), the strategist favored near-term dollar weakness (institutions are heavily underweight the dollar), while flagging dollar call options as "quite inexpensive": a late-summer buying opportunity on the view that US productivity ultimately supports the currency. The preferred trade: long the highest-yielding, least-crowded emerging-market currencies: "particularly Colombia, Brazil, and Mexico": funded by selling low-yielders like the Swiss franc and the Asian exporters.
  • A real-assets tilt: electrical infrastructure and Japanese industrials (analyst, positioning). Gromen, on Macro Voices (July 23), said US electrical-infrastructure equities are "in a really good place" because of power bottlenecks ("the United States has barely added any electrical capacity in 20 years"), pointing to the PAVE and GRID exchange-traded funds, and added Japanese industrial equities as a reshoring play.
  • A commodity contrarian: short the precious-metals miners, long copper (technician, operator). On The KE Report (July 23), Dana Lyons described holding short positions in gold and silver miners (covering the GLD near 353 and SLV near 48) while staying long copper (via the CPER fund), arguing gold needs "extended digestion" after its two-year run while copper, lacking a blow-off top, supports the longer-term bull case.

Read-throughs

  • The AI selloff is really a credit story, and it just reached a mega-cap. On Eurodollar University (July 22), the host walked through Amazon's latest bond sale, where investors demanded more generous pricing than usual: a signal that "Amazon's name and rating were no longer enough to produce limitless demand at any price." The scale of AI borrowing is the backdrop: AI-related debt issuance in 2026 has already reached roughly $270 billion, "nearly double the total issued during all of 2025," with Amazon, Alphabet, Meta, Microsoft and Oracle accounting for about $194 billion of it, and Amazon alone issuing about $107 billion over the past year. He tied it to PIMCO's warning that "the credit loss cycle is upon us," and to a slowdown in private lending (direct lending tied to buyouts fell "approximately 34% to $15.15 billion"). This connects straight to the stock story: Alphabet has raised more than $140 billion in debt and equity since October, and the podcasts noted its recent 100-year sterling bond is "down 7% since issuance." When the companies funding the boom start paying up to borrow, the boom's pace becomes hostage to the bond market's patience.
  • The 30-year Treasury is flashing a signal it hasn't in nearly two decades. As noted on Squawk on the Street (July 23), the 30-year yield has traded above 5% for 12 straight days: "the longest stretch since 2007." Coming alongside record corporate issuance to fund data centers and a globally synchronized move higher in long yields, it's the clearest sign that the cost of financing the AI buildout is rising in real time.
  • Immigration and demographics are quietly making the labor market almost impossible to break. This was the week's most important labor insight, and it came up repeatedly. Julia Coronado (above) argued that a "rapidly aging, retiring baby boom population" plus "extremely restrictive immigration policy… adds up to basically no labor force growth," which means "the unemployment rate really isn't going to rise unless the U.S. is in a recession." The State Street strategist on Street Signals put a number on it: the economy now needs only "somewhere between 0 and 50,000" new jobs a month to keep unemployment steady, while payrolls have actually grown an average of "92,000 per month this year": well above break-even, which is why "the unemployment rate ticked lower last month." Read-through: a resilient jobless rate is being propped up by a shrinking labor pool, not booming hiring, which supports the hawks (a tight labor market) and the demand-destruction camp (weak underlying momentum) at the same time.
  • A note on what was quiet this week: the labor data itself. The macro conversation was dominated by AI capex, oil, and the Fed; hard labor releases barely featured. The one fresh data point was weekly jobless claims: on One Rental At A Time (July 23), Michael Zuber flagged a new record low of "187,000… beating the old one of 189," and Squawk on the Street (July 23) noted the four-week average dropped to 207,000 from 214,000: "solid employment… good demand for labor." No one this week was invoking the Sahm rule (the recession-warning trigger based on a rising unemployment rate); with claims at record lows and the jobless rate falling, that particular alarm stayed silent.
  • Oil's new front: the Red Sea. The renewed energy spike isn't only about the Strait of Hormuz. On One Rental At A Time (July 23), Zuber flagged Houthi attacks on tankers in the Red Sea: a waterway that "controls 12% of global trade": stacking on top of the Hormuz risk, with Brent pushing toward triple digits. For the US, the read-through is inflation at the pump rather than 1970s-style shortages, but it's precisely the kind of supply shock that keeps the hawks from relaxing right before a Fed meeting.

What changed

Tuesday's recap left the Fed hawkish, the calendar sharpening, and the AI boom reframed as "a credit story." Five things moved since then, and they moved fast.

First, oil round-tripped. Tuesday we had WTI around $82–85 on Iran escalation; by Thursday the podcasts were talking about Brent back at $100 and a new Red Sea chokepoint, which reintroduced the inflation layer the disinflation camp thought was fading. Second, the AI-capex theme graduated from abstract to concrete and negative: where last week it was "spending will eat all of operating cash flow," this week Alphabet and Tesla actually reported, raised spending again, went cash-flow negative, and got sold: with Alphabet's first negative free cash flow in ~two decades and its bonds slipping. Third, the Fed split hardened into a genuine institutional fork: last week we described a "hawkish tone," but this week PGIM put a number on the hawkish case (three hikes) while Goldman and Morgan Stanley put an equally firm number on the dovish case (on hold, ~25% odds of a hike): a real disagreement among top desks, not a mood. Fourth, the K-shaped consumer story got its first serious, data-backed rebuttal (Adam Parker's "half reality, half overstatement"), turning a one-sided narrative into an actual debate. And fifth, a new wrinkle appeared that undercuts the bull case for the whole economy: several economists spent the week shrinking AI's estimated contribution to GDP: to roughly 40 basis points (Morgan Stanley) or as little as 10–30 basis points (Excess Returns), which means the economy is leaning on the consumer, and the consumer is leaning on the stock market.

The one thing that didn't change: the market and the Fed are still openly disagreeing about the next move. The difference is that this week the disagreement has both a hard date (Tuesday) and a fresh complication ($100 oil) attached to it.