Track Record

AMD

Apr 26, 2026 Insights

Pending

AMD is not a GPU company anymore. Data center crossed 48% of revenue in 2025 and is headed past 65% by 2027. The valuation re-rate happens through mix shift, not by beating NVIDIA. Intel is the real loser. The retail conversation is watching the wrong race.

Core Metric Tracking AMD data center segment revenue divided by total AMD revenue, full-year basis

Target Threshold Above 65%

Resolution Event AMD Q4 2027 earnings call

Review Date Feb 15, 2028

AMD

Apr 22, 2026 Earnings

Pending

AMD is becoming a genuine second AI platform, but only inside hyperscaler environments, and only if MI450 executes cleanly. The OpenAI deal is a deployment execution bet, not confirmed revenue.

Core Metric Tracking AMD data center segment annual revenue run rate AND number of publicly confirmed multi-year MI450 hyperscaler deployments

Target Threshold Data center AI revenue above $25 billion annualized AND at least two additional hyperscaler MI450 commitments confirmed publicly beyond current known deployments

Resolution Event AMD Q4 2026 earnings call

Review Date Feb 27, 2027

Original Analytical Thesis

Most coverage of AMD frames this as a NVIDIA competition story. It is not. NVIDIA has CUDA, better training performance, and a software ecosystem that took a decade to build. AMD is not catching that. What AMD is doing is carving out a parallel lane. The OpenAI deal is the clearest signal. OpenAI is not buying MI450 because H100 supply is tight. They are making a deliberate architectural commitment to a second platform for workload economics at scale. That is different from being a cheaper fallback. The market has not fully processed that distinction. AMD's stock still gets discussed as though its AI upside depends on NVIDIA stumbling. It does not. The hyperscaler "train on NVIDIA, infer on AMD" pattern is real and growing. Here is where I think the market is wrong: the OpenAI deal is being priced as confirmed revenue. It is a deployment execution bet. Six gigawatts of Instinct GPUs requires MI450 to ship on schedule, ROCm to work at scale, and Helios rack infrastructure to perform without the launch problems NVIDIA had in 2024. Any one of those slips and the revenue timeline moves right by two to three quarters minimum. The concentration risk is what the market is underweighting. AMD's entire AI revenue story sits inside a handful of hyperscalers. Three to five customers are the AI thesis. That is not a platform. That is customer concentration that introduces earnings volatility the current valuation does not fully reflect. One customer pause is a stock event. The inventory build makes this more pointed. AMD grew inventory from $5.7B to $7.9B year over year for MI450 demand. If the ramp slips or demand is overestimated, that becomes margin compression and write-down risk. My position: AMD is becoming a genuine second platform, but only inside hyperscaler environments, and only if MI450 executes cleanly

Live Monitoring Data

Primary Signal MI450 launch timing + hyperscaler MI450 commitment disclosures in H2 2026 earnings calls

Model Expectation Data center AI revenue annualizing above $25B by Q4 2026 + two or more new hyperscaler MI450 commitments

AMD

Apr 22, 2026 Macro

Pending

Sell-side consensus is mispricing AMD's hyperscaler concentration risk. Half of AMD's revenue depends on capex decisions made by five companies. When one pauses, it is a stock event, not a rounding error.

Core Metric Tracking AMD data center segment sequential revenue growth in Q2 and Q3 2026 + hyperscaler capex guidance language

Target Threshold If two or more of Microsoft, Meta, Google, Amazon shift to "optimizing" capex language AND AMD data center grows less than 5% sequentially in any quarter of 2026, the concentration risk thesis is confirmed

Resolution Event AMD Q3 2026 earnings call

Review Date Oct 26, 2026

AMD

Apr 10, 2026 Business Model

Pending

AMD is priced as an AI GPU company but reports in a way that makes that story unverifiable, and the valuation breaks the moment the market has to ask which business it was actually paying for.

Core Metric Tracking AMD Data Center segment YoY revenue growth rate - Q2 2026 (GAAP)

Target Threshold <20% YoY Data Center segment revenue growth (Q2 2026)

Resolution Event AMD Q2 2026 earnings call

Review Date Jul 31, 2026

Broadcom

Apr 7, 2026 Insights

Pending

Broadcom's XPU business generates at least $35 billion in cumulative AI semiconductor revenue across FY2026 and FY2027 combined - confirming that the hyperscaler design engagement model produces revenue at the scale Hock Tan described in December 2024, independent of whether the acquisition pipeline finds a next major target.

Core Metric Tracking Sum of Broadcom-reported AI semiconductor revenue across Q2, Q3, Q4 FY2026 and Q1, Q2 FY2027

Target Threshold >= $35 billion cumulative (floor) / >= $42 billion cumulative (confirmation of SAM trajectory)

Resolution Event Broadcom Q2 FY2027 earnings call

Review Date Jun 15, 2027

Broadcom

Apr 4, 2026 Macro

Pending

Hyperscaler AI capex remains growth-oriented through FY2026 - meaning Broadcom's AI semiconductor revenue does not fall below $9 billion in any quarter of FY2026 as long as Microsoft, Meta, and Google maintain their stated AI infrastructure buildout timelines.

Core Metric Tracking Broadcom AI semiconductor revenue remaining above $9 billion per quarter through Q4 FY2026

Target Threshold >=9.0B per quarter (floor) / consistent sequential growth (confirmation of trajectory)

Resolution Event Broadcom Q4 FY2026 earnings call

Review Date Dec 15, 2026

Broadcom

Apr 1, 2026 Earnings

Pending

VMware switching costs are high enough that FY2026 full-year infrastructure software revenue grows above 5%, confirming that retained customers paying higher subscription prices offset those who left post-acquisition.

Core Metric Tracking Broadcom Infrastructure Software segment revenue FY2026 full-year YoY growth (GAAP, as reported)

Target Threshold >=5.0% full-year YoY (floor) / >=10.0% (confirmation)

Resolution Event Broadcom Q4 FY2026 earnings call

Review Date Dec 12, 2026

Broadcom

Mar 29, 2026 Business Model

Pending

VMware's hypervisor lock-in is real enough that fewer than 20% of enterprise customers exit within the first full post-acquisition renewal cycle, confirming Broadcom's pricing power is structural, not contractual.

Core Metric Tracking Enterprise customer attrition rate in VMware infrastructure software segment across FY2026 full renewal cycle

Target Threshold <20% attrition = floor confirmation / <10% attrition = strong moat confirmed

Resolution Event Broadcom Q4 FY2026 earnings call

Review Date Dec 15, 2026

TSMC

Mar 26, 2026 Insights

Pending

TSMC N2 wafer output will reach or exceed 120,000 units per month by TSMC's Q4 2026 earnings call.

Core Metric Tracking N2 monthly wafer output (wafer starts per month)

Target Threshold ≥120,000 WSPM

Resolution Event TSMC Q4 2026 earnings call (expected January 2027)

Review Date Jan 31, 2027

TSMC

Mar 24, 2026 Macro

Pending

TSMC will report Q2 2026 gross margin at or above 62.0% confirming that N3 margin accretion and AI demand more than offset N2 ramp costs and overseas fab dilution.

Core Metric Tracking TSMC Q2 2026 gross margin (GAAP, USD terms)

Target Threshold ≥62.0%

Resolution Event TSMC Q2 2026 earnings call (expected July 2026)

Review Date Jul 31, 2026

TSMC

Mar 21, 2026 Earnings

Pending

N3 will be cited by TSMC management as having crossed the corporate gross margin average during the Q2 or Q3 2026 earnings call.

Core Metric Tracking TSMC management explicit confirmation of N3 crossing corporate gross margin average

Target Threshold Explicit statement in Q2 or Q3 2026 earnings call transcript

Resolution Event TSMC Q3 2026 earnings call (expected October 2026)

Review Date Oct 31, 2026

TSMC

Mar 18, 2026 Business Model

Pending

TSMC's FY2026 full-year gross margin will be at or above 61.0% confirming that $52-56B in CAPEX and overseas fab expansion does not erode the margin structure.

Core Metric Tracking TSMC FY2026 full-year gross margin (GAAP, USD terms)

Target Threshold ≥61.0%

Resolution Event TSMC Q4 2026 earnings call (expected January 2027)

Review Date Jan 31, 2027

Original Analytical Thesis

The market consistently undervalues TSMC relative to its fabless customers because foundries get framed as commodity manufacturers. TSMC is not a commodity manufacturer. Commodity manufacturers compete on price. TSMC competes on capability and no competitor has its capability at the leading edge. The $52-56 billion 2026 CAPEX is the number to anchor to. Not because it is large, but because TSMC commits it knowing it will compress gross margins by 2-4% in the near term from overseas fab dilution and N2 ramp costs. A commodity manufacturer does not accept margin compression to build infrastructure for 2028. A company with genuine conviction about multi-year demand does. C.C. Wei said on the Q4 2025 call that AI demand "looks like it is going to be endless." That is not a statement to take literally. It is a statement about confidence in the investment thesis behind $52-56 billion in capital allocation. Finovian's analytical stance: The single metric to track in 2026 is not revenue growth. It is N3 gross margin progression. Management said N3 will cross the corporate gross margin average in FY2026. N3 started commercial production in mid-2023 and was margin-dilutive for six quarters. If it crosses the corporate average in 2026, that confirms the pattern: new nodes start dilutive, scale to neutral, become accretive. N2 launched in Q4 2025. The clock on its progression has started. When N2 shows its first quarter at or above the corporate gross margin average that is the signal the margin expansion thesis is structurally intact for the next node cycle. Watch the per-node margin commentary in TSMC's quarterly management reports. That is where the real signal lives. This is an analytical observation, not investment advice. No price targets are expressed or implied.

Live Monitoring Data

Primary Signal Cumulative FY2026 gross margin tracking by Q2 and Q3. If running below 60% through H1, the full-year target requires a strong H2 recovery.

Model Expectation ≥61.0%

NVIDIA

Mar 15, 2026 Insights

Pending

NVIDIA Q1 FY2027 revenue will be at or above $76 billion, confirming hyperscaler AI CAPEX has not materially decelerated entering 2026.

Core Metric Tracking NVIDIA Q1 FY2027 reported revenue

Target Threshold ≥$76.0 billion

Resolution Event NVIDIA Q1 FY2027 earnings call (expected May 2026)

Review Date May 31, 2026

NVIDIA

Mar 13, 2026 Macro

Pending

At least one of Microsoft or Amazon will guide sequential AI CAPEX growth below 10% QoQ in their Q2 2026 earnings — the first signal that the AI infrastructure arms race is transitioning toward a returns-driven phase.

Core Metric Tracking Sequential AI infrastructure CAPEX growth guidance from Microsoft or Amazon

Target Threshold <10% sequential QoQ CAPEX growth guidance from either company

Resolution Event Amazon Q2 2026 earnings OR Microsoft Q4 FY2026 earnings (both expected July 2026)

Review Date May 31, 2026

NVIDIA

Mar 10, 2026 Earnings

Pending

NVIDIA Q1 FY2027 gross margin will be at or above 74.0% confirming Blackwell scaling has not produced pricing concessions that signal competitive pressure reaching NVIDIA's margin structure.

Core Metric Tracking NVIDIA Q1 FY2027 GAAP gross margin

Target Threshold ≥74.0%

Resolution Event NVIDIA Q1 FY2027 earnings call (expected May 2026)

Review Date May 31, 2026

NVIDIA

Mar 8, 2026 Business Model

Pending

Google and Amazon will not cite internal custom silicon displacing more than a minor share of external GPU procurement in any 2026 earnings call — confirming CUDA switching costs hold the competitive moat intact through FY2026.

Core Metric Tracking Explicit management commentary from Google or Amazon citing meaningful TPU/Trainium displacement of NVIDIA GPU procurement

Target Threshold Neither company cites material displacement in any 2026 earnings call

Resolution Event Amazon Q4 2026 earnings (expected February 2027) — full-year 2026 coverage

Review Date Feb 28, 2027

AMD is not a GPU company anymore. Data center crossed 48% of revenue in 2025 and is headed past 65% by 2027. The valuation re-rate happens through mix shift, not by beating NVIDIA. Intel is the real loser. The retail conversation is watching the wrong race.

Original Analytical Thesis

Live Monitoring Data

AMD is becoming a genuine second AI platform, but only inside hyperscaler environments, and only if MI450 executes cleanly. The OpenAI deal is a deployment execution bet, not confirmed revenue.

Original Analytical Thesis

Live Monitoring Data

Sell-side consensus is mispricing AMD's hyperscaler concentration risk. Half of AMD's revenue depends on capex decisions made by five companies. When one pauses, it is a stock event, not a rounding error.

Original Analytical Thesis

Live Monitoring Data

AMD is priced as an AI GPU company but reports in a way that makes that story unverifiable, and the valuation breaks the moment the market has to ask which business it was actually paying for.

Original Analytical Thesis

Live Monitoring Data

Original Analytical Thesis

Live Monitoring Data

Hyperscaler AI capex remains growth-oriented through FY2026 - meaning Broadcom's AI semiconductor revenue does not fall below $9 billion in any quarter of FY2026 as long as Microsoft, Meta, and Google maintain their stated AI infrastructure buildout timelines.

Original Analytical Thesis

Live Monitoring Data

VMware switching costs are high enough that FY2026 full-year infrastructure software revenue grows above 5%, confirming that retained customers paying higher subscription prices offset those who left post-acquisition.

Original Analytical Thesis

Live Monitoring Data

VMware's hypervisor lock-in is real enough that fewer than 20% of enterprise customers exit within the first full post-acquisition renewal cycle, confirming Broadcom's pricing power is structural, not contractual.

Original Analytical Thesis

Live Monitoring Data

TSMC N2 wafer output will reach or exceed 120,000 units per month by TSMC's Q4 2026 earnings call.

Original Analytical Thesis

Live Monitoring Data

TSMC will report Q2 2026 gross margin at or above 62.0% confirming that N3 margin accretion and AI demand more than offset N2 ramp costs and overseas fab dilution.

Original Analytical Thesis

Live Monitoring Data

N3 will be cited by TSMC management as having crossed the corporate gross margin average during the Q2 or Q3 2026 earnings call.

Original Analytical Thesis

Live Monitoring Data

TSMC's FY2026 full-year gross margin will be at or above 61.0% confirming that $52-56B in CAPEX and overseas fab expansion does not erode the margin structure.

Original Analytical Thesis

Live Monitoring Data

NVIDIA Q1 FY2027 revenue will be at or above $76 billion, confirming hyperscaler AI CAPEX has not materially decelerated entering 2026.

Original Analytical Thesis

Live Monitoring Data

At least one of Microsoft or Amazon will guide sequential AI CAPEX growth below 10% QoQ in their Q2 2026 earnings — the first signal that the AI infrastructure arms race is transitioning toward a returns-driven phase.

Original Analytical Thesis

Live Monitoring Data

NVIDIA Q1 FY2027 gross margin will be at or above 74.0% confirming Blackwell scaling has not produced pricing concessions that signal competitive pressure reaching NVIDIA's margin structure.

Original Analytical Thesis

Live Monitoring Data

Google and Amazon will not cite internal custom silicon displacing more than a minor share of external GPU procurement in any 2026 earnings call — confirming CUDA switching costs hold the competitive moat intact through FY2026.

Original Analytical Thesis

Live Monitoring Data

Never Miss a Shift