Compare moats
Up to three covered companies, band by band. Every call is a curated editorial judgment, never a disclosed figure — and every band carries its cited basis.
| Cipher Mining | Core Scientific | ChipMOS Technologies | |
|---|---|---|---|
| Moat rating | none Cipher (renamed Cipher Digital Inc. on 20 February 2026) holds signed, credit-supported leases but has not yet shown that they amount to a durable advantage. Its 2025 Form 10-K (filed 2026-02-24) says "Through the end of 2025, our revenue has been derived from mining bitcoin". Its Q2 2026 business update (2026-08-04, Exhibit 99.1, https://www.sec.gov/Archives/edgar/data/1819989/000181998926000038/q226_earningsxprxdraftxvf.htm) reported "Q2 2026 Revenue of $25 million", and its first HPC rent began only in August 2026 at Black Pearl. The 10-K's own risk factors describe a contested market. They say "There has been an increasing number of businesses constructing HPC data centers, which has resulted in increasing competition and pricing pressure that may cause us to reduce our pricing in order to remain competitive", that "our data centers are single-tenant properties", and that lessees "will have the right to terminate the lease if there are significant delays in the completion of construction". At Barber Lake, where the 10-K targeted Phase I delivery "by September 30, 2026", a September 2026 amendment, made "In connection with change orders and the continued evolution of tenant requirements", moved data-hall deliveries to the fourth quarter of 2026 through the first quarter of 2027. Under it, Cipher "will bear the first $359.3 million of costs in excess of the initial budgeted amount" (2026-09-25). There is real counter-evidence: a 15-year Amazon lease, a Google-backstopped Fluidstack lease whose contracted life a "leading AI lab" extended to 20 years, and a third lease with an investment-grade hyperscale tenant. Until that capacity is delivered and paying rent, though, a moat is not shown, so the band is none. | narrow Core Scientific's position rests on long-dated, take-or-pay leases of energized capacity, not on a protected market. Its 2025 Form 10-K (filed 2026-03-02) says wholesale colocation "typically involves large, long-term agreements with a limited number of customers, often with initial terms of 10 years or more". It adds that such leases are frequently paired with take-or-pay commitments "under which the customer is obligated to pay for leased customer power capacity regardless of utilization, providing operators with revenue visibility over the contract term." On the Q2 2026 call (2026-07-28, https://d1io3yog0oux5.cloudfront.net/_fda3a1f2f1890adaa7ac6c4d9d310971/corescientific/db/1085/11280/prepared_remarks/2Q26+Prepared+Remarks+Transcript+PDF.pdf) management said the company "now has approximately 1.1 gigawatts of total contracted billable capacity, representing more than $24 billion of base contracted revenue". That total combines CoreWeave's 590 MW, on what the CFO called "the 12-year lease terms", with AMD agreements for 530 MW that carry "more than $14 billion of base contracted revenue across the 15-year agreements with 2.5% annual escalators". Management also said it was "currently billing for 437 megawatts of capacity". The limits are just as clear. The 10-K says "Competitors compete on price, facility location, reputation and perceived skill with respect to performance" and that many rivals "are more established, have better brand recognition, are well capitalized". It also says "One customer, CoreWeave, currently accounts for 100% of our Colocation segment revenue." The Q2 2026 10-Q says a material weakness tied to converting mining facilities to HPC "has not been remediated". Long take-or-pay terms are the norm in wholesale colocation, so the contracts alone are contracted revenue, not an edge over rivals. What supports narrow is capacity already billing and a tenant that has expanded in place, and even that is bound to those contracts and to one customer today, so the band is narrow, not wide. | none ChipMOS earns thin returns in businesses its own 20-F describes as price-driven. The FY2025 20-F reports gross margin of 10.8% in 2025, against 13.0% in 2024. Profit attributable to equity holders was NT$551 million, against NT$1,440 million in 2024 and NT$1,968 million in 2023. The display driver business, 24.5% of 2025 revenue, saw its gross margin fall to 7.5% from 22.5%, which the filing attributes to 'the decline of average selling price'. Its revenue fell 20% on customer inventory adjustments and 'the price competition among DDIC suppliers in Mainland China'. Assembly services ran at a -1.6% gross margin. The filing says prices for its services 'tend to decline over the course of its product and technology life cycle'. Results improved in the 2026 memory upturn: the 2Q26 press release reports an 18% gross margin and the 2Q26 call reports a 14.4% return on equity. One strong quarter in a memory upturn does not offset profits that fell in both 2024 and 2025, so the rating is none. |
| Moat type | none No moat source is demonstrated yet. Switching costs are the most likely candidate, since the leases are long (15 years at Black Pearl; Barber Lake now a 20-year contracted life) and each site has a single tenant. But the 10-K says tenant guarantees "will only be effective after rent commencement under such leases and are subject to certain limitations", and by August 2026 rent had begun at only one site. The advantages the 10-K claims are the company's own. It speaks of "industry-leading expertise in originating and securing industrial-scale, greenfield data center sites" and of securing West Texas land "on more favorable terms than in more established data center markets". The same document says "there is significant competition for power capacity and energized facilities". Intellectual property is modest: "four granted United States patents and one issued patent in Taiwan". With 66 full-time employees, Cipher has no scale advantage over the competitors it names: CoreWeave, Digital Realty, Equinix, Vantage Data Centers and Aligned Data Centers. | switching costs The lock-in is contractual and physical. The 10-K's description of colocation is that "Customers deploy and manage their own hardware, while the operator designs, builds and operates the underlying infrastructure", under take-or-pay leases "often with initial terms of 10 years or more". On the Q2 2026 call the COO described the AMD program as "an integrated colocation platform engineered to support AMD's Helios Rack-Scale systems" with a "close-coupled AMD design framework". Leaving would mean rebuilding that power, cooling and fit-out elsewhere. The relationship with CoreWeave shows the stickiness: it "began with a 16-megawatt lease at our Austin campus in 2024 and has since expanded to 590 megawatts of total contracted capacity". Intellectual property is not the source. The 10-K says the company has "filed over 130 patent applications" but that these laws and procedures "provide only limited protection". Scale is not the source either: the 10-K says rivals are "well capitalized" and some are organized to lower "their external cost of capital". | none None of the moat sources is supported. Switching costs are low: the 20-F says customer qualification 'typically takes up to eight weeks', that test correlation 'typically takes up to two weeks', and that most customers buy through purchase orders with 'three-month non-binding rolling forecasts'. Scale sits elsewhere: TrendForce's 2024 ranking (2025-05-13) places ChipMOS tenth among OSATs at $710M, against $2.28B for Powertech, which TrendForce (2026-01-12) calls the world's leading DRAM and NAND test and packaging provider. IP is not decisive: the 20-F lists the measures of competitiveness in testing as software engineering, quality of service, flexibility, capacity, production cycle time and price. It names ASE, Amkor, Chipbond, KYEC, Powertech, JCET and UTAC as competitors, alongside IDMs with in-house capacity. |
| Leadership | fast follower No independent share or rank was found. The 10-K names CoreWeave, Digital Realty, Equinix, Vantage Data Centers and Aligned Data Centers as competitors, along with miners that have "signed leases with hyperscalers and HPC tenants": Hut 8, IREN, TeraWulf, Core Scientific and Applied Digital. Cipher has signed hyperscale-grade tenants (Amazon; Fluidstack with a Google backstop; a third investment-grade hyperscaler) but delivered its first HPC capacity only in August 2026 (Q2 2026 update). That makes it a fast follower. Its self-description as "a leading developer, owner, and operator of industrial-scale data centers" is the company's own claim and is not counted. | fast follower No independent share or rank was found. The 10-K lists Aligned Data Centers, Compass Datacenters, Equinix, Digital Realty Trust, NTT, QTS, Switch, Vantage Data Centers and CyrusOne as HDC competitors and says "Many of these competitors are more established, have better brand recognition, are well capitalized". It separately names Applied Digital, Cipher Mining, Galaxy Digital, Hut 8, IREN and TeraWulf as miners converting facilities for AI and HPC customers. Against that field Core Scientific is scaling quickly from a small base: 437 MW billing and about 1.1 GW contracted at the Q2 2026 call. That makes it a fast follower behind larger incumbents. Its press releases describe the company as "a leader in designing, building and operating large scale, purpose-built data centers", but that is the company's own claim and is not counted. | at parity TrendForce's 2024 ranking (2025-05-13) places ChipMOS tenth among global OSATs with $710M in revenue, and credits its driver IC business with 'steady demand in the automotive and OLED sectors'. Powertech's 2025 annual report, compiled from Taiwan's Market Observation Post System, shows ChipMOS revenue of NT$23,933 million in 2025, against NT$21,454 million for Chipbond, a competitor the 20-F names, and NT$74,929 million for Powertech. The 20-F calls ChipMOS 'one of the leading independent providers' of display driver and memory testing and assembly in Taiwan, which is its own description. No source shows a lead over peers, so the band is at_parity. |
| Pricing power | weak The 10-K says competition "has resulted in increasing competition and pricing pressure that may cause us to reduce our pricing in order to remain competitive". It adds that if customers cut usage "we may be compelled to lower our prices or risk losing a significant customer". At Barber Lake, Cipher agreed to absorb "the first $359.3 million of costs in excess of the initial budgeted amount", with the tenant reimbursing "50% of any such costs above that amount" (2026-09-25, Exhibit 99.1, https://www.sec.gov/Archives/edgar/data/1819989/000181998926000043/barberlakeleaseamendmentpr.htm). Its one input-cost edge is specific to mining: Odessa's power costs about 2.8 c/kWh under the Luminant contract, available "until at least July 2027". A landlord that concedes cost overruns to its tenants has weak pricing power. | moderate Price is set competitively at signing and then locked in. The 10-K says "Competitors compete on price" and warns that "If we fail to accurately estimate the factors upon which we base our contract pricing, we may generate less profit than expected or incur losses on those contracts". Once signed, the AMD agreements carry "2.5% annual escalators" (Q2 2026 call), and the Q2 2026 10-Q says "power costs are passed through to our customer without markup". To win AMD, the company also gave AMD "market-priced warrants to purchase Core Scientific's common stock, subject to certain commercial conditions" (AMD release, 2026-07-28, https://investors.corescientific.com/news-events/press-releases/detail/138/core-scientific-and-amd-announce-infrastructure-partnership). In Q2 2026, colocation revenue was $136,669 thousand and the cost of colocation services was $56,686 thousand, including $35,073 thousand of power fees passed through to the customer (10-Q). Contracted escalators protect price over a 12- to 15-year term, but nothing shows a premium over rivals, so pricing power is moderate. | weak The 20-F says its service fees 'tend to decline in tandem with the declining average selling prices of the products we test and assemble'. It says memory fees 'were sharply reduced in tandem with' falling DRAM and flash prices in the past. Display driver gross margin fell to 7.5% from 22.5% in 2025 on falling selling prices. In the 2026 upturn the company is passing costs through: on the 2Q26 call the chairman said rising costs for substrates, lead frames and gold will be passed on to customers, and gross margin reached 18.0%. That is cost recovery in a tight market, not durable pricing power. |
| Summary | Cipher built bitcoin mining data centres in Texas and is now developing single-tenant AI and HPC campuses for lease to hyperscalers. Its 10-K reports a portfolio of "4.2 gigawatts ("GW") of capacity across 10 sites". It lists a 15-year Amazon Web Services lease for about 300 MW of turnkey capacity at Black Pearl and a Fluidstack lease at Barber Lake (300 MW gross) under which Google "has agreed to backstop certain obligations of Fluidstack". It describes bitcoin mining at Odessa on a Luminant power contract at about 2.8 c/kWh. In 2026 it signed a third campus lease "with an investment-grade Hyperscale tenant" (Q1 2026 update, 2026-05-05), delivered first Black Pearl capacity in August "two months ahead of the original schedule" with rent commenced, and fully funded its Stingray development with a bond (Q2 2026 update). On 2026-09-25 it said Barber Lake's contracted life was extended from 10 to 20 years, taking contracted revenue at the site "from $3.8 billion to over $9 billion". The 10-K says Odessa was "the first bitcoin mining data center awarded the Management and Operations, or M&O, Stamp of Approval award from the Uptime Institute", which is independent recognition of how the company operates. Against this, the latest quarter's revenue was still bitcoin mining and fell to $25 million. The 10-K describes growing "competition and pricing pressure", single-tenant concentration, and termination rights for construction delays. The Barber Lake schedule was reset, and Cipher absorbs the first $359.3 million of cost overruns. Cipher's contracted pipeline is substantial, but a competitive advantage is not yet demonstrated, so it is rated as having no moat. That could change to narrow once its leased campuses are delivered and paying rent. | Core Scientific began as a bitcoin miner and is turning its sites into high-density colocation (HDC) data centres for AI and HPC tenants. At 31 December 2025 it owned or leased ten data centres across seven U.S. states, with "approximately 1.4 gigawatts ("GW") of gross utility power capacity, or approximately 920 megawatts ("MW") of total leasable customer power capacity", and it intends "to convert every megawatt in our portfolio to high-density colocation infrastructure over the next three years" (2025 10-K). The transition is now visible in the numbers: colocation revenue "represented 77% of total revenue" in the first half of 2026, against 12% a year earlier (Q2 2026 10-Q). By July 2026 the company was billing for 437 MW. It had about 1.1 GW contracted with two anchor customers, CoreWeave at 590 MW and AMD at 530 MW (about 380 MW leased directly to AMD and about 150 MW for a neocloud that AMD backs), and AMD holds an exclusive right, under specified conditions, to lease as much as 2 additional gigawatts (Q2 2026 call). Those long take-or-pay contracts, plus energized power at a time when the 10-K says utilities demand "significant collateral postings at contract execution" and large builds take "18 to 24 months, or longer", are the source of a narrow advantage. They do not make a protected franchise. The 10-K names nine established colocation providers and six converted miners as competitors, says competitors "compete on price", and reports that CoreWeave was 100 percent of colocation revenue. The Q2 2026 10-Q says colocation revenue "is concentrated with a single customer" and that the conversion-related material weakness is still unremediated. Build costs are "approximately $11 million to $12 million per megawatt" (Q2 2026 call), and the company expects to finance the AMD build-out through project-level bonds. Core Scientific is rated as having a narrow moat: contract-bound switching costs, with no evidence of a price premium or a scale advantage. | ChipMOS is a Taiwanese outsourced assembly and test house (OSAT) with plants in Hsinchu and Tainan. In 2025, memory and logic/mixed-signal testing was 23.7% of revenue, assembly 28.6%, display driver assembly and test 24.5% and bumping 23.2%. Its top five customers took 61% of revenue, and its top fifteen include Micron, Novatek, Himax, Nanya, Winbond, Macronix and MediaTek. The memory upturn has shifted the mix. On the 2Q26 call, memory was 51% of revenue and display drivers 18.1%, overall utilization was 72%, and management said it 'selectively raised the memory OSAT price' to reflect material costs. 2Q26 revenue of NT$7,383.1 million was the highest since 2014. TrendForce (2026-01-12) cites analysts saying that DRAM, 'led by DDR4, which makes up roughly 70–80% of memory sales', remains the company's 'core earnings driver'. Management is reallocating capacity. It signed a three-year memory wafer test take-or-pay contract in early 2026, it is 'monetizing the low-end and lower UT level DDIC assets', and it plans to add testing capacity for AI ASICs and silicon photonics. The verdict is no moat: a capable mid-sized OSAT whose margins follow memory prices and Chinese display-driver competition. |
| Chain position | Cipher develops single-tenant, powered data-centre campuses for hyperscale and AI tenants: Amazon at Black Pearl, Fluidstack (backstopped by Google) at Barber Lake, and an unnamed investment-grade hyperscaler at a third campus. Meanwhile it winds down bitcoin mining. | Core Scientific leases powered, liquid-cooled data-centre capacity to AI compute providers: CoreWeave (590 MW) and AMD, including a neocloud that AMD backs (about 530 MW), per the Q2 2026 call. It sits between utilities and GPU cloud operators while it winds down bitcoin self-mining. | Back-end assembly and test supplier to memory makers and display-driver designers. Customers headquartered in Taiwan were 87% of 2025 revenue, per the 20-F. |
| Products (share / barrier) |
|
|
|
| Long-horizon vote | -0.06 at weight 0.20 · swarm bearish Editorial prior, not backtested. | +0.06 at weight 0.20 · swarm bullish Editorial prior, not backtested. | -0.06 at weight 0.20 · swarm bearish Editorial prior, not backtested. |