Muse AI agent led Meta Platforms into the Best Stocks in the Market list. What Josh Brown thinks

Meta shares have risen 38% from their August $545 low to about $750 following the September 8 launch of Muse, while the company forecasts Q3 revenue of $61B-$64B and 2026 capex of $130B-$145B. The article argues Muse validates Meta's AI investment and increases demand for AMD hardware, including up to 6GW of MI450 GPUs and Venice CPUs. AMD is up 194% year-to-date at roughly $631 after reaching a $1T market capitalization; Q2 revenue rose 50% to $11.5B, Data Center revenue jumped 107% to $6.7B, and Q3 revenue guidance is approximately $13B.

Analysis

The key underwriting change is that Meta's AI spend can now be valued as a retention, engagement and transaction-intent engine—not merely as infrastructure expense. The highest-value monetization path is likely not a standalone subscription: better targeting, higher advertiser conversion and commerce/referral economics can create operating leverage across the existing ad base. Over the next 1-3 quarters, the market will demand evidence in time spent, query/task frequency, ad load and cost-per-action; absent those KPIs, incremental capex is still a multiple risk despite the product narrative.

AMD's opportunity is materially larger than a single customer win, but the economics need scrutiny. Large capacity commitments and equity-linked warrants can pull forward reported demand while transferring upside to hyperscalers and potentially dilute per-share returns; the critical missing data are contractual take-or-pay terms, deployment timing, realized rack utilization and AMD gross-margin progression after custom-system mix. AVGO is the more direct competitive read-through on custom silicon, while NVDA retains the software-ecosystem and supply-chain advantage; a broad shift toward heterogeneous agent inference would benefit server CPU suppliers but may not automatically imply AMD share gains.

Near term, both equities are momentum-sensitive and vulnerable to a positioning reset rather than a fundamental break. A 1-3 month pullback is most likely if Meta's next results show expense acceleration ahead of monetization or if AMD's backlog converts slower than investors assume; over 6-18 months, proof that agent workloads increase inference intensity would expand the total compute market rather than simply reallocate GPU spend. Contrarian view: the market may be over-crediting consumer adoption for immediate revenue while under-crediting the longer-run strategic value of proprietary behavioral data and advertiser distribution if usage persists.

Key Decisions for Investors

• Do not chase META at current momentum levels; establish a 1-3 month long only on a successful retest of the prior breakout area near $685, with a closing stop below $685 for a tactical position. Upside requires next-quarter evidence of AI-driven engagement or ad-conversion improvement; a weekly loss of the 200-day area near $626 invalidates the structural setup.
• Use AMD as a conditional long, not a fresh full-size momentum entry: buy a $600-$610 support hold or sell cash-secured puts at/around the $575 strike for 1-2 month expiry. Stop a directional long on a close below $575; upside is a continuation toward new highs if deployment/backlog commentary confirms revenue conversion, while the primary risk is custom-hardware margin dilution or delayed installations.
• Express the compute-architecture uncertainty through a smaller pair: long AMD / short AVGO only after AMD holds $600 and only into the next deployment or earnings catalyst. This isolates a potential merchant GPU-and-CPU share-gain thesis from general AI beta; exit if AMD breaks $575 or if management discloses weaker gross-margin trajectory despite revenue growth.
• Maintain NVDA exposure rather than rotating wholesale into AMD: treat any evidence of lower-than-expected inference utilization or customer deployment delays as sector-level risk, not proof of share transfer. Watch hyperscaler capex guidance and server-lead-time commentary over the next two reporting cycles before increasing semiconductor gross exposure.