Bernstein comments on Coherent (COHR.US) Q1 performance: no problem in itself; the focus is on when to deliver and maintain the “outperform the market” rating

The Zhitong Finance App learned that Bernstein released the latest research report to judge the Coherent (COHR.US) quarterly results: there is no problem with the quarterly report itself; the question is “when will it be delivered.” Coherent's revenue for the first quarter of fiscal year 2027 was $2.37 billion and earnings per share of $2.09, exceeding the consensus by 2% and 4%, respectively; the FY2027 guideline was a 60% increase in revenue, which was about 10 percentage points higher than the consensus. But the three levers underpinning the $350 target — the first $4 billion in single-quarter revenue, the first 45% gross margin quarter, and scale-up optical emissions — all fall from the end of 2027 to 2028, or fiscal year 2028. The report maintains ratings and target prices, and points out that there is no clear catalyst before the March OFC meeting.

First quarter: data center single leg support

Total revenue of $2.37 billion is basically the same as Bernstein's $2,372 million, and surpassed the consensus of $2,315 million by about 2%. Two faces: data centers and communications at $1,952 million, surpassing Bernstein's $1.88 billion by 4%; industry, at $419 million, is 16% lower than Bernstein's 497 million and 8% lower than the agreed $453 million. Data centers and communications already account for 82% of revenue, and industrial weakness is completely covered by AI optical communication — this is why Bernstein doesn't overturn judgments with industrial mistakes.

Excess profit margins are due to expenses, and cash flow is flawed

Non-GAAP gross margin of 40.6%, lower than Bernstein's 41.0%, same as consensus; operating margin of 23.2%, surpassing Bernstein's 20.3% and consensus 22.8%; earnings per share of $2.09, surpassing Bernstein's 2.05 and consensus 2.02. The gross margin was unsurprising and the operating profit margin clearly exceeded expectations, indicating that the excess was due to cost control rather than product structure or capacity efficiency — this corresponds exactly to what investors have been dissatisfied with for a long time.

The real flaw lies in the cash flow: free cash flow was negative $868 million, and Bernstein's original forecast was negative 617 million, but the consensus forecast was positive 286 million. The market originally expected to generate cash this season; in fact, there was a huge outflow. Bernstein is relatively forgiving: the second half of the fiscal year is expected to be positive, and the full year of fiscal year 2028 is positive; after adding debt in the first quarter, debt is still only 1.35 times the annually-adjusted EBITDA; long-term agreements include advance payments prior to capital expenditure, so capital demand is moderate. In an environment where equity financing is generally heavily penalized, it is more realistic to rely on debt to resolve incremental capital.

Guidelines: 60% growth and two “firsts”

Second-quarter guidance revenue of $2.55 billion to $2.75 billion, earnings per share of $2.28 to $2.48, gross profit margin of 40% to 42%, non-GAAP operating expenses of $430 million to $450 million, and tax rates of 18% to 20%. According to the median value, revenue was 2% higher than Bernstein's original forecast and 4% higher than consensus; earnings per share were 4% higher than Bernstein's, and 6% higher than consensus.

The full-year guidance is even more impactful: revenue for the 2027 fiscal year increased by 60%, corresponding to about US$11.3 billion, higher than Bernstein's original forecast of 10.8 billion dollars and the consensus of 10.7 billion dollars. The company gave two “firsts” at the same time — the first quarterly revenue of $4 billion in fiscal year 2028 and the first 45% gross margin quarter before the end of the 2028 natural year.

The second promise is worth recounting. Starting with the median value of 41% of the second quarter guidance, if 45% falls in the fourth quarter of the 2028 fiscal year (corresponding to the second quarter of fiscal year 2029), the middle 8 quarters would need to increase by an average of about 50 basis points per quarter, which is slightly faster than the actual 40 basis points of the first quarter and the second quarter guidance, but it is close to the trajectory of the previous 8 quarters; superimposed product structure and internal indium phosphide (InP) production capacity climbing. Bernstein believes it is “believable.” Based on this, it raised its forecast: earnings per share for the second quarter rose to 2.45, with full-year revenue of $11.5 billion and earnings per share of $11.12, up 16.9% from the previous $9.51.

CPO math: 7.7 times optical content, $16.2 billion opportunity in 2028

The most informative part of the report was a verifiable calculation of Nvidia's Rubin Ultra's optical usage. Coherent revealed at the PhotonLink event that a 100 Tbps input/output computing or switching chip with 200 Gbps (500 channels) per channel corresponds to a content value of approximately $15,000, which is equivalent to $30 per channel.

According to the design given by Nvidia in OCP: 8 racks are a scale-up unit, 72 GPUs per rack, and a two-tier network; the inner leaf node of the rack is a 144-port NVLink 7 co-packaged optical (CPO) exchange tray (2 200G channels per port), 72 ports connected to the rack GPU and 72 ports to the ridge node; the ridge side is estimated to have a 4-ridge structure, and each spine link is bundled with 18 ports. Bernstein assumed that the GPU would use a copper cable to the leaf, a CPO from the leaf to the top of the ridge, and a pluggable module on the ridge side. According to this, each plane has a CPO bandwidth of 230.4 Tbps, which is pluggable; Nvidia also disclosed that each Rubin GPU has an NVLink bandwidth of 3 Tb/s (12 Tbps, 30 400G ports), bringing 30 planes, for a total of 13,824 Tbps.

Keep the scale-out of the NVL72 at 32.4 Tb/s, and the 8-rack unit is 259.2 Tbps. The result: 8-rack optical bandwidth increased from 2,073.6 Tbps (full scale-out) to 15,897.6 Tbps (scale-out 2,073.6, scale-up pluggable 6,912, scale-up CPO 6,912), and optical content grew 7.7-fold; in linear terms of value, a single 8-rack unit was approximately $2.4 million.

Multiply by shipment volume: Bernstein's Asia Semiconductor team (Mark Li) estimates about 1 million R300 units in 2027 and 3.9 million units in 2028, corresponding to approximately 1,800 and 6,800 Rubin Ultra units (576 GPUs per unit). At $2.4 million per unit, the opportunities are around $4.2 billion in 2027 and $16.2 billion in 2028. Bernstein's own hypothesis was 4 million in 2028, and the conclusion was unanimous.

The weakest premise of this set of mathematics is that the price per Gbps remains the same — historically, this price has continued to decline, and the current relative advantage is due to InP supply bottlenecks. Coherent indicates that InP production capacity will double by the end of 2027; even if Rubin Ultra only accounts for 10% of Nvidia's total GPU shipments, doubling demand is not aggressive; VCSEL and MicroLED will separate part of the demand during the same period, and Bernstein still judges that supply and demand will remain favorable until 2027. It retains the observation point that other players in the ecosystem are climbing the slope — Chinese manufacturers such as AXTI, Sumitomo, and Yuanjie are not covered by it.

$20 billion long term and “likely to be the target of acquisition”

The company emphasized increased visibility, expanded coverage of long-term agreements, and for the first time presented a specific scale: a $20 billion long-term agreement, lasting 5 years, starting in early 2028. This is the second major approximate length after supplying Nvidia's CPO with a 400 milliwatt continuous wave laser. Bernstein's attitude is that “progress is exciting, but the historical record is shaky” — using long-term contracts to hedge cycle risks, past success rates have not been high.

Regarding the merger between Marvell and Lumentum, the report gives a counterintuitive observation: major mergers and acquisitions often benefit competitors. Mergers focus within the delivery and integration period, and competitors can spread uncertainty and concerns among customers. Coherent also used this to emphasize that it does not compete with customers, and that Marvell has competitive relationships with many Lumentum customers. The report further speculates that Coherent itself may also be a more attractive acquisition target, particularly for customers concerned about supply; management emphasizes satisfaction with the current trajectory.

There is also an unkind hint in the same paragraph: Bernstein “fears Lumentum may sell the business before fundamentals deteriorate”, and if so, it is equally negative for Coherent; it also points out that management “doesn't care, may be too displeased” about the threat to Asian competitors.

Valuation and Ratings

The target price of $350 corresponds to two equivalent paths: 31 times FY2027 earnings of $11.12 per share, or 19 times FY2028 earnings of $18.04 per share. Based on the closing price of $19.5.09 on November 4, there was an implied increase of 79%; the market capitalization was US$38.425 billion, and the corporate value was US$40.311 billion.

The difference in expectations is mainly at the far end: earnings of $11.12 per share for FY2027 were 17% higher than consensus 9.50, FY2028 $18.04 was 26% higher than consensus 14.30, and FY2029 $28.38 was 43% higher than consensus 19.81. There are four risks: the popularity of CW+SiPh lasers alleviates InP bottlenecks and weakens pricing power; China uses InP export restrictions to counter US tariffs on optical modules; CPO yield issues lag behind optical adoption; and AI investment slows down.

The stock price has already dropped one round: 42.1% in the past month and 40.9% in the past six months, losing 41.7 and 47.6 percentage points, respectively, compared to the S&P 500. While acknowledging short-term catalytic shortages, Bernstein gave a 79% upside — a contradiction that requires readers to balance themselves.