AI energy infrastructure buildout: Valero Energy (VLO) vs. Marathon Petroleum (MPC)

McKinsey estimates that approximately $5.2 trillion of capital investment may be required globally by 2030 to support AI-related data-center demand. Its model assumes AI-related data-center capacity reaches approximately 156 GW by 2030, requiring about 125 GW of incremental capacity between 2025 and 2030. Of that $5.2 trillion, approximately $1.3 trillion is expected to go to the “energizers”—power generation and transmission, cooling, electrical equipment and related infrastructure.

That enormous investment requirement has led me to look beyond semiconductor companies and hyperscalers at the physical infrastructure required to support the AI buildout. Power generation, natural gas, electrical equipment, cooling, construction and other infrastructure are increasingly part of the AI investment story.

An interesting signal recently emerged from my own sector-performance dashboard.

The Dashboard Identified Something Unexpected

I maintain a screen of major companies across all 11 S&P 500 industry sectors. The dashboard measures approximately 12-month price performance and supplements that with Bollinger Band position, moving-average position, RSI and other technical indicators.

When I reviewed the highest-performing stocks, several Energy companies appeared near the top. More interestingly, the standout performers were not primarily the integrated oil companies.

They were refiners.

Valero Energy (VLO), Marathon Petroleum (MPC) and Phillips 66 (PSX) have dramatically outperformed not only the S&P 500, but also the broader Energy sector and the other major companies in the Energy screen.

At a high level, the performance hierarchy looks approximately like this:

The individual-stock figures are approximately 12-month price returns from the BlueByrd Portfolio Dashboard. The S&P 500 and XLE figures are broad reference comparisons; exact returns vary with measurement date and whether dividends are included.

The progression is striking:

S&P 500 ~15–19% → Energy ~46% → major refiners ~111–148%.

But the comparison becomes even more interesting when we look inside the Energy sector.

Refiners Stand Out Within the Energy Sector

The dashboard tracks the ten major holdings in the Energy universe. Ranked by approximately 12-month performance, they currently look like this:

Source: BlueByrd Portfolio Dashboard. Approximately 12-month price returns.

The dispersion is remarkable.

All three major refiners in the group—Valero, Marathon Petroleum and Phillips 66—have generated returns exceeding 100%. The next-highest company, SLB, is at 50.7%.

In other words, the three refiners have generated more than twice the 12-month return of every other company in the Energy Top 10.

This suggests that the performance cannot be explained simply by investors rotating into Energy. There has been a much more pronounced rerating of the refining companies.

That observation led to the question behind this article:

Why have VLO and MPC performed so extraordinarily well, and how much of that performance can actually be attributed to the AI-energy infrastructure buildout?

The Immediate Explanation Is Refining Economics—not AI

The first conclusion from the research is important.

The extraordinary performance of VLO and MPC should not primarily be attributed to AI.

Both companies have benefited from a highly favorable refining environment, including much stronger crack spreads, constrained refining capacity, resilient transportation-fuel demand and geopolitical disruption.

Their second-quarter 2026 results illustrate just how significant the change has been.

Valero reported Q2 net income of $3.7 billion, or $12.62 per share, compared with only $714 million, or $2.28 per share, one year earlier. Its Refining segment generated $4.5 billion of operating income, compared with $1.3 billion in Q2 2025. Valero also returned $2.6 billion to shareholders during the quarter.

Marathon’s results were similarly dramatic. MPC reported Q2 net income of $5.1 billion, or $17.73 per diluted share, compared with $1.2 billion, or $3.96 per share, a year earlier. Adjusted EBITDA increased from $3.3 billion to $8.5 billion, while the company returned $2.8 billion of capital to shareholders.

The key number may be Marathon’s refining margin.

MPC’s Refining & Marketing margin increased from $17.58 per barrel in Q2 2025 to $36.33 in Q2 2026. Refining & Marketing adjusted EBITDA increased from $1.9 billion to $6.7 billion, and management said the results were driven primarily by higher crack spreads across all regions.

This distinction matters.

The dashboard identified VLO and MPC because their stocks had generated extraordinary returns. The subsequent research indicates that the refining cycle—not AI infrastructure—is the principal explanation for those returns to date.

VLO and MPC: Strong Similarities, One Important Difference

At first glance, VLO and MPC look remarkably similar.

Both operate enormous refining systems. Both are currently generating exceptional earnings and cash flow. Both are returning large amounts of capital to shareholders.

Their recent numbers illustrate the similarities:

The difference that is particularly interesting in the context of AI infrastructure is MPLX.

Marathon’s MPLX Stake Creates an AI-Infrastructure Option

Marathon Petroleum’s most tangible connection to the AI infrastructure buildout is through MPLX, its midstream business.

That distinction is important.

MPC itself is not an AI company, and AI workloads are not responsible for the extraordinary refining earnings Marathon is currently reporting. But MPLX owns and operates natural-gas processing, gathering, pipeline and related infrastructure that could become increasingly valuable as data-center developers search for large quantities of reliable power.

The collaboration between MPLX and MARA Holdings provides a concrete example.

In November 2025, MPLX and MARA announced a letter of intent under which MPLX would facilitate natural-gas supply from its Delaware Basin processing facilities to planned integrated power-generation facilities and data-center campuses in West Texas.

The initial planned capacity is approximately 400 MW, with the potential to scale to 1.5 GW.

MARA describes the strategy as combining access to natural gas with on-site power generation and compute infrastructure. The planned sites are intended to integrate power generation directly with large computing loads rather than relying exclusively on conventional grid expansion.

That model is particularly interesting because power availability is becoming one of the constraints on AI infrastructure development.

McKinsey estimates that U.S. data-center electricity consumption could increase from approximately 3–4% of total U.S. power demand to roughly 11–12% by 2030.

MPLX therefore provides Marathon with something Valero does not have to the same degree: an existing infrastructure platform that could participate in increasing natural-gas demand associated with power generation and data-center development.

I would characterize that as strategic optionality, rather than current AI earnings.

AI isn’t what produced MPC’s $5.1 billion quarter. Refining economics did.

But if data-center power demand develops as expected, MPLX gives MPC another potential avenue for growth.

Valero Is a Different Investment Thesis

Valero does not have an equivalent midstream platform providing the same obvious pathway into natural-gas-powered data-center infrastructure.

Its investment thesis is therefore more directly tied to refining economics.

That is not necessarily a disadvantage.

Valero’s Refining segment generated $4.5 billion of operating income during Q2, while throughput averaged approximately 3.0 million barrels per day. Its Renewable Diesel segment also improved dramatically, generating $717 million of operating income, compared with a $79 million loss a year earlier.

Valero therefore offers relatively direct exposure to the economics that have actually driven the refinery rally.

The distinction can be summarized simply:

VLO is primarily a refining-cycle investment.

MPC combines refining-cycle exposure with a substantial midstream business that provides additional natural-gas and potentially data-center infrastructure exposure.

That makes the two companies interesting for somewhat different reasons despite their superficially similar businesses.

What the Dashboard Says About the Stocks Today

Strong fundamentals do not necessarily imply an attractive entry price.

That is where the second part of the dashboard becomes useful.

The dashboard does not attempt to determine intrinsic value. Instead, it measures where the current stock price sits relative to its recent statistical trading range.

The latest Energy screen looks like this:

Source: BlueByrd Portfolio Dashboard. %b represents the stock’s position relative to its current Bollinger Bands; 0 corresponds approximately to the lower band and 1 to the upper band.

This produces another striking separation.

All seven non-refiners in the Energy Top 10 currently have %b readings of 0.28 or lower. The three refiners are at 0.68 to 0.72.

Valero currently trades at $393.27 against dashboard Bollinger Bands of approximately $327.29 and $419.45, placing it at about 72% of the range.

Marathon trades at $402.38 against bands of approximately $345.17 and $429.47, placing it at about 68%.

Phillips 66 is similarly positioned at approximately 68%.

The important point is not that a higher %b makes these companies fundamentally expensive. It doesn’t. Bollinger Bands measure price behavior, not intrinsic value.

Instead, the dashboard is telling us that the three companies that have produced the most extraordinary long-term performance also remain considerably higher within their recent statistical price ranges than the rest of the Energy group.

That is useful information when thinking about entry timing.

For me, it separates two different questions:

Is this a company I want to own?

and

Is this the price at which I want to add to it?

Fundamental research addresses the first question. The dashboard provides additional evidence for the second.

Risks

The largest risk to both VLO and MPC is that today’s exceptional refining economics prove temporary.

Marathon’s Q2 Refining & Marketing margin of $36.33 per barrel was more than twice the $17.58 reported one year earlier. Refining & Marketing adjusted EBITDA increased from $1.9 billion to $6.7 billion.

Valero experienced a similar expansion, with Refining operating income increasing from $1.3 billion to $4.5 billion.

Those results demonstrate the enormous earnings power of the businesses under favorable conditions. They also illustrate their cyclicality.

If crack spreads normalize materially, earnings can decline rapidly even if the companies continue to operate their refineries effectively. A low earnings multiple calculated using unusually high cyclical earnings can therefore be misleading.

Geopolitical conditions are another important variable. Disruptions to crude-oil and refined-product markets can create unusually favorable regional margins for refiners. A normalization of those conditions could remove part of the premium currently embedded in refining economics.

There is also price risk after the rally itself. Valero has gained approximately 148%, Marathon approximately 127% and Phillips 66 approximately 111% over the dashboard’s measurement period. Even fundamentally strong companies can experience substantial corrections after moves of that magnitude.

The dashboard provides some evidence of that risk. Despite recent declines, VLO, MPC and PSX remain considerably higher within their Bollinger ranges than the other major Energy holdings.

MPC has an additional thesis-specific risk: investors should not assume that data-center demand will automatically translate into material Marathon earnings.

The MPLX/MARA initiative is tangible, and a potential 1.5 GW deployment would be significant. But the agreement began as a letter of intent, and the eventual economics will depend on project execution, capital requirements, customer commitments, utilization and the actual cash flows generated.

More broadly, the AI infrastructure forecasts themselves contain uncertainty. McKinsey’s $5.2 trillion estimate is a forecast based on assumptions about AI-related data-center demand, infrastructure requirements and deployment through 2030—not committed spending.

AI demand may continue to expand rapidly while the timing, location and beneficiaries of the infrastructure investment differ substantially from today’s expectations.

Takeaways

The most useful insight from this analysis was not the one I expected when I started.

The dashboard initially identified VLO and MPC because their performance was extraordinary. Energy had substantially outperformed the broader market, but VLO, MPC and PSX had in turn dramatically outperformed the Energy sector itself.

Looking at all ten major Energy holdings made that observation much clearer.

The three refiners returned approximately 111% to 148% over the dashboard’s measurement period. Every other company in the Energy Top 10 returned 51% or less.

Researching why produced a more nuanced conclusion.

First, the refinery rally is supported by actual earnings. Valero and Marathon have experienced enormous increases in refining profitability and cash generation. These are not simply stocks caught up in an AI narrative.

Second, AI does not explain most of those gains. The immediate driver has been refining economics, particularly much stronger margins and crack spreads.

Third, MPC has an additional dimension that VLO largely lacks. Through MPLX, Marathon has substantial natural-gas and NGL infrastructure that could participate in the rapidly growing demand for power associated with data centers. The MPLX/MARA initiative provides a concrete example of how that opportunity could develop.

Fourth, VLO and MPC therefore represent related but different investment propositions. VLO provides more direct exposure to the refining cycle. MPC combines refining exposure with a large midstream platform and an emerging data-center infrastructure opportunity.

Finally, price still matters. VLO and MPC have already appreciated approximately 148% and 127%, respectively. Their current %b readings of 0.72 and 0.68 are well below the upper Bollinger Band, but they remain substantially higher within their recent price ranges than the other major Energy holdings.

That is why I find combining systematic screening with fundamental research useful.

The screen tells me where something unusual is happening.

Fundamental research helps determine why it is happening and whether it may be sustainable.

And price-position indicators provide another piece of information when deciding when the risk/reward may be more attractive.

In this case, the screen led from an unexpected group of market leaders to a broader conclusion: the refinery rally is principally a refining-cycle story today, while MPC’s MPLX platform provides an intriguing additional route into the longer-term AI energy-infrastructure buildout.