Japanese SaaS Hits Its 'Common AI Socket' Moment: freee Group's Logikura and freee Inventory Management Announce MCP Servers the Same Day, With Six Official MCP Announcements Linked by This Site in Two Days -- From 'Done by You' to 'Done for You'
ANK-Doc ID: ANK-2026-06-30-002 Version: v1.0.0 Published: 2026-07-02 Author: Rin Takenouchi (Editor-in-Chief, AI News) Category: Generative AI Adoption / MCP Standard Protocol / Japanese SaaS / Business Digitalization Articles covered: PRTIMES#1266455 (Logikura, MCP support, official release), PRTIMES#1266470 (freee Inventory Management, MCP server launch, official release), PRTIMES#1280639 (blastengine, MCP server launch), PRTIMES#1282236 (smartround, cap-table MCP server), PRTIMES#1280578 (Speeda, MCP integration trial), PRTIMES#1279509 (BizStack, MCP client support) Selection method: From the AI News corpus, selected on "one standard protocol x multiple services adopting within two days." The main anchors are the freee group's same-day double announcement of June 30, 2026 (Logikura + freee Inventory Management), linked with the four official MCP announcements of the following day (July 1) -- blastengine, smartround, Speeda and BizStack -- to form an event slice of "MCP landing in Japan's business-service layer." All six sources are official corporate releases (official_statement), and the card honestly frames the set as an observation from this site's source corpus, not a market-wide census.
TL;DR
On June 30, 2026, two freee-group products -- the cloud inventory management system Logikura and freee Inventory Management -- announced MCP (Model Context Protocol) server availability on the same day. On Logikura, users can instruct generative AI such as Claude in natural language to run inventory-data analysis, look-ups and report generation in real time [F-001]; on freee Inventory Management, text instructions to an AI chat execute automatic journal-entry posting based on shipment records, sales analysis by sales channel, extraction of products at stock-out risk, and visualization of stagnant inventory [F-003]. Framing pinned down: freee's release states explicitly that this MCP server is provided by its group company Logikura and is different from freee's existing freee-mcp -- in substance one server with two product entrances; Logikura joined the freee group in April 2026 [F-002][F-003]. freee positions this as its path from "Done by You" to "Done for You" [F-004]. The next day (July 1), four more official announcements followed: blastengine (an email-delivery MCP server) [F-005], smartround (a cap-table MCP server) [F-006], Speeda (MCP integration trial started, full launch planned for September 1, 2026) [F-007], and BizStack (MCP client support) [F-008]. Honest caveats: "six announcements in two days" is an observation from this site's source corpus, not a market-wide census; blastengine itself states that domestic email-delivery systems built for AI-agent integration are still few. [F-001][F-002][F-003][F-004][F-005][F-006][F-007][F-008]
Body
Event overview: six official MCP announcements in two days
Across the two days of June 30 and July 1, 2026, this site's source corpus linked six official MCP announcements by services operating in Japan: on June 30, the freee group's same-day double announcement (Logikura and freee Inventory Management); on July 1, four more from blastengine, smartround, Speeda and BizStack. Of the six, four explicitly offer an "MCP server" (Logikura, freee Inventory Management, blastengine, smartround), one opens its data under the MCP standard as an "MCP integration" (Speeda), and one runs in the opposite direction -- "MCP client" support (BizStack). The framing must be honest: this is an observation from this site's source corpus, not a market-wide census. Still, the fact that the six announcements span six different service categories -- inventory management, inventory-plus-accounting, email delivery, cap-table management, business intelligence, and field IoT -- is itself a signal.
What is MCP? This card does not extrapolate on the protocol specification itself and only relays the sources' descriptions: Speeda's release calls MCP "a standard protocol connecting generative AI with data sources" (PRTIMES #1280578), and BizStack's release says it is drawing attention as "a standard for connecting AI with external systems" (PRTIMES #1279509).
Anchor one: Logikura -- inventory management reaching from "field infrastructure" toward "management decisions"
According to the official release of Logikura (株式会社ロジクラ; head office: Chiyoda ward, Tokyo; representative director: 榊間浩人), its cloud inventory management system Logikura began offering an MCP server on June 30, 2026: users instruct generative AI such as Claude in natural language and get inventory-data analysis, look-ups and report generation executed in real time; Logikura has been used by a cumulative total of more than 30 thousand users (PRTIMES #1266455). [F-001]
The release lists three conversational experiences. First, instant grasp of sales by channel -- asking the AI which channel grew last month returns a sales ranking and composition ratios based on shipment records. Second, early detection of stock-out risk -- automatic matching against replenishment-threshold inventory lists the products to reorder in priority order. Third, visualization and monetization of stagnant inventory -- stock with no shipments over a set period is grasped instantly by quantity and appraised value (PRTIMES #1266455). The company's representative director (榊間浩人) commented that Logikura is to evolve from a tool that "records and manages inventory" into a foundation that "supports management decisions."
On security architecture, the release states explicitly: this is not a structure in which external AI accesses the database directly; Logikura's own system intermediates, so operations continue under the security policies the field has built up (PRTIMES #1266455). [F-002] This is the company's own description of its architecture, and this card makes no independent security assessment.
Anchor two: freee Inventory Management -- same-day announcement, same server
The same day, freee K.K. (フリー株式会社; head office: Shinagawa ward, Tokyo; CEO: 佐々木大輔) announced that freee Inventory Management began offering a remote MCP server: text instructions to an AI chat suffice to run automatic journal-entry posting based on shipment records, sales analysis by sales channel, extraction of products at stock-out risk, and visualization of stagnant inventory (PRTIMES #1266470). [F-003] The background the release describes: at many small and medium-sized companies and e-commerce operators, sales/inventory management and accounting systems are split, and channel-level sales tallies and stock-out/stagnation checks depend on manual spreadsheet work, hindering timely management decisions.
The framing must be pinned down: freee's release states explicitly that the MCP server offered here is provided by its group company Logikura and is different from freee's existing freee-mcp (using both together allows information to be used across the products). Moreover, the "how to use" sections of both releases point to the same "Logikura remote MCP server" setup document -- in substance, one server with two product entrances (PRTIMES #1266470, PRTIMES #1266455). Reading the "same-day double announcement" as two independent MCP servers would be an over-extension of the framing.
The group background of the double announcement: Logikura joined the freee group in April 2026, states in its release that its development pace has since accelerated, and shipped this MCP support following its freee-accounting integration release (PRTIMES #1266455). [F-002] freee, for its part, positions this in its June 30, 2026 release as its D4U path from "how easy it is for humans to use (Done by You)" to "how fully it can be entrusted to AI (Done for You)," framed by 4 values -- low-cost design, a sustainable growth model, high safety standards, and reliable execution (PRTIMES #1266470). [F-004] These are the two companies' own statements of product philosophy, not verifiable results.
Same-period spread: email delivery, cap tables, business intelligence, field IoT
The next day, July 1, 2026, at least four more official MCP announcements followed.
blastengine (email delivery): blastengine, the email-delivery system of Rakus Light Cloud (株式会社ラクスライトクラウド; head office: Shibuya ward, Tokyo), a subsidiary of Rakus (株式会社ラクス), began offering the "blastengine MCP server" from July 1, 2026: natural-language instructions from AI tools such as Claude or Cursor execute email delivery and the checking and analysis of delivery results. The release adds that overseas, led by major email-delivery providers, an environment where AI agents use email-delivery services autonomously is taking shape, while domestic email-delivery systems built for AI-agent integration are still few (PRTIMES #1280639). [F-005] That "still few in Japan" is the company's own market observation, not quantified in the source -- but it honestly marks MCP support in Japan as still early-stage.
smartround (cap tables): smartround (株式会社スマートラウンド; head office: Chiyoda ward, Tokyo; representative director: 砂川大) announced its "cap-table MCP server" and "cap-table creation Skills": startup founders and staff can reference and operate the cap-table data stored in smartround directly from general-purpose AI tools such as Claude or ChatGPT, creating, simulating and revising cap tables through dialogue; the background problem it describes is the "triple wall" of specialized knowledge, operational burden, and the absence of someone to consult (PRTIMES #1282236). [F-006]
Speeda (business intelligence): "Speeda MCP Integration," from the business-intelligence platform Speeda, started trial availability on July 1, with full availability planned for September 1, 2026. It opens Speeda's company, industry, financial and expert-insight data under the MCP standard so users can call it directly from the generative AI tools they use daily; the release says Speeda's accumulated data is used in the decision-making of more than 3,000 companies, and points to the challenge of "accuracy, expertise and currency" of information in AI use (PRTIMES #1280578). [F-007] Note: the full-launch date is a plan, not an accomplished fact.
BizStack (field IoT): MODE, Inc., a Silicon Valley startup (head office: California, USA; Japan branch: Chiyoda ward, Tokyo; CEO: 上田学), announced MCP client support for its field IoT platform BizStack: the AI assistant "BizStack Assistant" can now connect with the data sources and external systems held by user companies. The release notes that until now, connecting AI with business systems required custom development company by company, a barrier to adoption (PRTIMES #1279509). [F-008] The direction must be kept distinct: the other five open their own services' data to external AI via MCP (server side / data opening), while BizStack has its own AI assistant reach out to external data as an MCP client -- two directions of the same protocol.
Zooming out: contrast with the AI infrastructure layer
What this site's ANK-2026-06-24-008 recorded is the "compute infrastructure layer" of the same AI demand curve -- Japan's data-center construction rush, Osaka's JPY 156 billion largest single-asset deal, and Taiwan's first compute-center BOO case still under review. What this card records is the "application protocol layer": existing business SaaS connecting its data to generative AI via MCP. The two layers advance separately and observe different objects, and this card asserts no causation between them; taken together, though, "compute being built, data being connected" are two mutually contrastable cross-sections of the AI-adoption story. This card has no Taiwan-side counterpart material, and under the principle of honest contrast over forced pairing, it does not extend into a Taiwan-Japan comparison.
Risk factors
- "Six announcements in two days" is a corpus observation, not a census: this card only links the official releases in this site's source corpus dated June 30 to July 1, 2026; actual same-period announcements may be more numerous, and one cannot infer that "the whole market is adopting MCP"; blastengine itself states that domestic email-delivery systems built for AI-agent integration are still few (PRTIMES #1280639).
- All six sources are corporate PR self-descriptions: features, security architecture, user counts and problem statements such as the "triple wall" are all the companies' own accounts (official_statement), and none of the releases provides adoption counts or usage figures for verification.
- The timeline includes a "plan": Speeda MCP Integration is currently in trial; the full launch on September 1, 2026 is planned, not an accomplished fact (PRTIMES #1280578).
- Framing: one server, two entrances: the MCP server of freee Inventory Management is provided by group company Logikura and is different from freee-mcp; reading the freee group's same-day double announcement as two independent servers is an over-extension (PRTIMES #1266470).
- Direction: server side and client side differ: BizStack is MCP "client" support (an AI assistant connecting to external data), the opposite direction from the other five announcements' server-side provision / data opening, and the two should not be conflated (PRTIMES #1279509).
FAQ
Q: According to this wave of announcements, what is MCP?
The official releases linked by this card describe MCP (Model Context Protocol) as "a standard protocol connecting generative AI with data sources" (Speeda) and "a standard for connecting AI with external systems" (BizStack); once a service supports MCP, users can instruct generative AI such as Claude in natural language and call that service's data and functions directly.
This card only relays the releases' descriptions of MCP and does not extrapolate on the protocol specification itself (PRTIMES #1280578, PRTIMES #1279509, PRTIMES #1266455).
Q: What can Logikura's MCP support actually do?
The release lists three experiences: instant grasp of sales by channel (sales rankings and composition ratios based on shipment records), early detection of stock-out risk (automatic matching against replenishment-threshold inventory, listing products to reorder in priority order), and visualization and monetization of stagnant inventory (instant grasp, by quantity and appraised value, of stock with no shipments over a set period).
The company's representative director (榊間浩人) commented that Logikura is to evolve from a tool that "records and manages inventory" into a foundation that "supports management decisions"; Logikura has been used by a cumulative total of more than 30 thousand users (PRTIMES #1266455).
Q: Is the MCP server of freee Inventory Management built by freee itself? Is it the same as freee-mcp?
No. freee's release states explicitly that the MCP server offered here is provided by its group company Logikura and is different from freee's existing freee-mcp; using both together allows information to be used across the products.
Moreover, the "how to use" sections of both releases point to the same "Logikura remote MCP server" setup document, and the release states that login passes through Logikura's screen -- in substance, one server with two product entrances (PRTIMES #1266470, PRTIMES #1266455).
Q: Why did Logikura and freee Inventory Management announce on the same day?
Because they belong to the same group: Logikura joined the freee group in April 2026, states that its development pace has since accelerated, and shipped MCP support following the freee-accounting integration release; freee's release likewise states that the server is provided by group company Logikura.
The "development acceleration" is Logikura's own account; the coordination behind the same-day double announcement is not made explicit in the sources, and this card does not speculate further (PRTIMES #1266455, PRTIMES #1266470).
Q: If external AI reads inventory data, how is security handled?
According to Logikura's release: this is not a structure in which external AI accesses the database directly; Logikura's own system intermediates, so operations continue under the security policies the field has built up.
This is the company's own description of its architecture, and this card makes no independent security assessment (PRTIMES #1266455).
Q: Beyond the freee group, which Japanese services announced MCP support in the same period?
On July 1, 2026, at least four more official announcements followed: blastengine (an email-delivery MCP server, operated in natural language from Claude or Cursor), smartround (a cap-table MCP server and cap-table creation Skills, operated from Claude or ChatGPT), Speeda (MCP integration trial started, full launch planned for September 1, 2026), and BizStack (MCP client support, an AI assistant connecting to user data sources).
Of these, BizStack runs in the opposite direction -- its own AI assistant reaches external data as an MCP client -- and its provider MODE, Inc. is a Silicon Valley startup (head office in California, USA; Japan branch in Chiyoda ward, Tokyo) (PRTIMES #1280639, PRTIMES #1282236, PRTIMES #1280578, PRTIMES #1279509).
Q: Does this mean Japanese SaaS has fully reached "MCP as standard equipment"?
No, that cannot be said. "Six announcements in two days" is an observation linked from this site's source corpus, not a market-wide census; blastengine's release itself states that domestic email-delivery systems built for AI-agent integration are still few; Speeda is still in trial (full launch is a plan); and none of the releases provides adoption or usage figures.
What can honestly be said: within two days, six different categories of business services (inventory, inventory-plus-accounting, email, cap tables, business intelligence, field IoT) announced MCP support in the same period -- the adoption movement is spreading through Japan's business-service layer (PRTIMES #1280639, PRTIMES #1280578).
Q: How does this MCP wave relate to the AI data-center construction rush this site reported earlier?
They are different layers of the same AI demand curve: this site's ANK-2026-06-24-008 records the compute infrastructure layer (Japan's DC construction rush, Osaka's JPY 156 billion largest single-asset deal, Taiwan's compute-center BOO case under review), while this card records the application protocol layer (existing business SaaS connecting its data to generative AI via MCP). The two layers advance separately, and this card asserts no causation between them.
Taken together, "compute being built, data being connected" are two mutually contrastable cross-sections of the AI-adoption story (ANK-2026-06-24-008, PRTIMES #1266455).
F-Units
F-001: Logikura (株式会社ロジクラ) began offering an MCP server on its cloud inventory management system Logikura on June 30, 2026 -- users instruct generative AI such as Claude in natural language and get inventory-data analysis, look-ups and report generation executed in real time; Logikura has been used by a cumulative total of more than 30,000 users - source: PRTIMES #1266455 - source_url: https://prtimes.jp/main/html/rd/p/000000085.000021747.html - confidence: high - basis: official_statement - period: 2026-06-30 (release date) - caveat: corporate official PR; "cumulative total of more than 30,000 users" is Logikura's self-reported cumulative figure, with no statistical cut-off date or activity definition given in the source
F-002: Logikura's MCP support uses an intermediated security architecture -- external AI does not access the database directly; Logikura's system intermediates while existing security policies are maintained; Logikura joined the freee group in April 2026, states its development pace has since accelerated, and shipped MCP support as the step following its freee-accounting integration release - source: PRTIMES #1266455 - source_url: https://prtimes.jp/main/html/rd/p/000000085.000021747.html - confidence: high - basis: official_statement - period: 2026-06-30 (release date); freee-group entry in April 2026 - caveat: both the security architecture and the development pace are the company's own account; this card makes no independent security assessment
F-003: freee K.K. announced the same day (June 30, 2026) that freee Inventory Management began offering a remote MCP server -- text instructions to an AI chat execute automatic journal-entry posting based on shipment records, sales analysis by sales channel, extraction of products at stock-out risk, and visualization of stagnant inventory; the MCP server is provided by group company Logikura and is different from freee's existing freee-mcp - source: PRTIMES #1266470 - source_url: https://prtimes.jp/main/html/rd/p/000002105.000006428.html - confidence: high - basis: official_statement - period: 2026-06-30 (release date) - caveat: feature descriptions are freee's official PR self-description; the release states that login passes through Logikura's screen
F-004: freee positions this as its D4U path from "how easy it is for humans to use (Done by You)" to "how fully it can be entrusted to AI (Done for You)," with D4U framed by 4 values (low-cost design, a sustainable growth model, high safety standards, reliable execution) - source: PRTIMES #1266470 - source_url: https://prtimes.jp/main/html/rd/p/000002105.000006428.html - confidence: medium - basis: official_statement - period: 2026-06-30 (release date) - caveat: a statement of freee's product philosophy, not a verifiable performance metric
F-005: blastengine, the email-delivery system of Rakus Light Cloud (株式会社ラクスライトクラウド), a subsidiary of Rakus (株式会社ラクス), offers the "blastengine MCP server" from July 1, 2026 -- natural-language instructions from AI tools such as Claude or Cursor execute email delivery and the checking and analysis of delivery results; the release states that overseas, led by major email-delivery providers, autonomous AI-agent use of email delivery is taking shape, while domestic email-delivery systems built for AI-agent integration are still few - source: PRTIMES #1280639 - source_url: https://prtimes.jp/main/html/rd/p/000000190.000017058.html - confidence: high - basis: official_statement - period: 2026-07-01 (availability start date) - caveat: "still few in Japan" is the company's own market observation in the release, not quantified in the source
F-006: smartround (株式会社スマートラウンド) announced its "cap-table MCP server" and "cap-table creation Skills" -- startup founders and staff can reference and operate the cap-table data stored in smartround directly from general-purpose AI tools such as Claude or ChatGPT, creating, simulating and revising cap tables through dialogue - source: PRTIMES #1282236 - source_url: https://prtimes.jp/main/html/rd/p/000000092.000042542.html - confidence: high - basis: official_statement - period: 2026-07-01 (announcement) - caveat: service description is the official PR's self-description; the "triple wall" (specialized knowledge, operational burden, absence of someone to consult) is the background problem as stated by the company
F-007: "Speeda MCP Integration," from the business-intelligence platform Speeda, started trial availability on July 1, 2026, with full availability planned for September 1, 2026 -- it opens Speeda's company, industry, financial and expert-insight data under the MCP standard for direct calls from the generative AI tools users use daily; the release states Speeda's accumulated data is used in the decision-making of more than 3,000 companies - source: PRTIMES #1280578 - source_url: https://prtimes.jp/main/html/rd/p/000000280.000010548.html - confidence: high - basis: official_statement - period: 2026-07-01 (trial start); full availability planned for 2026-09-01 - caveat: "more than 3,000 companies" is Speeda's self-reported figure; the full-launch date is a plan, not an accomplished fact
F-008: MODE, Inc., a Silicon Valley startup (head office: California, USA; Japan branch: Chiyoda ward, Tokyo), announced MCP client support for its field IoT platform BizStack -- the AI assistant "BizStack Assistant" can connect with the data sources and external systems held by user companies; the release notes that connecting AI with business systems previously required custom development company by company, a barrier to adoption, and that MCP is drawing attention as "a standard for connecting AI with external systems" - source: PRTIMES #1279509 - source_url: https://prtimes.jp/main/html/rd/p/000000247.000035514.html - confidence: high - basis: official_statement - period: 2026-07-01 (release date) - caveat: MODE's official PR self-description; BizStack is MCP "client" support, a direction different from the other five announcements' server-side provision / data opening
J-Units
J-001: Across the two days of June 30 to July 1, 2026, the MCP announcements linked by this site's source corpus reached six, spanning inventory management, inventory-plus-accounting, email delivery, cap tables, business intelligence and field IoT -- indicating that in Japan's business-service layer MCP is moving from case-by-case integration toward standard-protocol support; but this is a corpus observation, not a market-wide census, blastengine itself says domestic support is still scarce, Speeda is still in trial, and "standard equipment" status cannot yet be claimed - confidence: medium - basis: inferred
J-002: The freee group's "same-day double announcement" is in substance one server with two product entrances -- the MCP server of freee Inventory Management is provided by group company Logikura, and both releases point to the same "Logikura remote MCP server" setup document; after Logikura joined the freee group in April 2026, the freee-accounting integration and MCP support shipped in succession, and the post-merger development cadence is the backdrop of the same-day announcement (the acceleration claim is Logikura's own) - confidence: medium - basis: inferred
J-003: The rhetoric shared by the six releases is "open the existing service's data to the generative AI users already use," not "build our own AI" -- 3 of the 6 releases name Claude explicitly (Logikura, blastengine, smartround), and they variously appeal to security design (Logikura's system-intermediated model) and data quality (the accuracy/expertise/currency challenge Speeda cites); this is an observation of the shared points of the release texts, and none of the sources provides actual adoption or usage results - confidence: medium - basis: inferred
P-Units
P-001: Actual adoption and usage of each service's MCP support -- none of the six releases provides usage counts or adopting-company numbers, so whether "support" converts into "use" must be verified against each company's future disclosures ### P-002: Whether "Speeda MCP Integration" reaches full availability as planned on September 1, 2026 -- it is currently in trial and the full launch is a plan; the announcement at that time needs tracking ### P-003: Whether MCP support spreads from these six service categories to more Japanese business SaaS -- blastengine's own observation that domestic AI-agent-ready systems are still few awaits market-side verification, including the moves of established large SaaS vendors
同事件・三視角 / Three Perspectives on the Same Event / 同一イベント・三つの視点
Internal Citation Chain
Published ANK-Docs cited by this article: - ANK-2026-06-24-008 (Taiwan-Japan AI data-center build-out: Taiwan's first compute-center BOO case still under review vs Osaka's JPY 156 billion "largest single-asset" deal) -> cited as the contrast on the "compute infrastructure layer" of the same AI demand curve -- that card records compute and data centers being built, while this card records the application protocol layer (existing SaaS connecting its data to generative AI via MCP); the two layers advance separately, and this card asserts no causation between them.
Sources
1. [PRTIMES #1266455] Logikura (株式会社ロジクラ), "From foundation to compass: Logikura supports the AI standard MCP (Model Context Protocol)" (土台から、羅針盤へ。ロジクラ、AI標準規格「MCP（Model Context Protocol）」に対応), 2026-06-30. https://prtimes.jp/main/html/rd/p/000000085.000021747.html 2. [PRTIMES #1266470] freee K.K., "freee Inventory Management begins offering an MCP server" (「freee 在庫管理」、MCPサーバーを提供開始), 2026-06-30. https://prtimes.jp/main/html/rd/p/000002105.000006428.html 3. [PRTIMES #1280639] Rakus Light Cloud (株式会社ラクスライトクラウド, Rakus group), "Email-delivery system blastengine launches an MCP server enabling AI-agent integration" (メール配信システム「blastengine」、AIエージェント連携を実現する「MCPサーバー」を提供開始), 2026-07-01. https://prtimes.jp/main/html/rd/p/000000190.000017058.html 4. [PRTIMES #1282236] smartround (株式会社スマートラウンド), "smartround releases a cap-table MCP server letting startups consult on and build cap tables from Claude or ChatGPT" (スマートラウンド、スタートアップがClaudeやChatGPTから資本政策の相談・作成ができる「資本政策MCPサーバー」をリリース), 2026-07-01. https://prtimes.jp/main/html/rd/p/000000092.000042542.html 5. [PRTIMES #1280578] Speeda, "Speeda launches the new service 'Speeda MCP Integration,' making business intelligence callable from major generative AIs" (Speeda、新サービス「Speeda MCP連携」の提供を開始。主要な生成AIから経済情報の呼び出しが可能に), 2026-07-01. https://prtimes.jp/main/html/rd/p/000000280.000010548.html 6. [PRTIMES #1279509] MODE, Inc., "BizStack supports MCP client, strengthening 'AI that understands the field'" (BizStack、MCPクライアントに対応。"現場を理解するAI"を強化), 2026-07-01. https://prtimes.jp/main/html/rd/p/000000247.000035514.html 7. [ANK-2026-06-24-008] Rin Takenouchi, "Taiwan-Japan AI data-center build-out: Taiwan's first compute-center BOO case is still under review (NT$300m investment floor, 15 PetaFLOPS compute floor) vs Osaka's JPY 156 billion 'largest single-asset' deal, with JLL projecting Japan's DC market at US$33.4 billion by 2030", 2026-06-28. https://ainews.washinmura.jp/ainews/en/ank/ANK-2026-06-24-008